Development and Application of Software Tools for Mass ...

ACTAUNIVERSITATIS

UPSALIENSISUPPSALA

2017

Digital Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Pharmacy 234

Development and Applicationof Software Tools for MassSpectrometry Imaging

PATRIK KÄLLBACK

ISSN 1651-6192ISBN 978-91-513-0040-5urn:nbn:se:uu:diva-328016

Dissertation presented at Uppsala University to be publicly examined in A1:111a, BMC,Husargatan 3, Uppsala, Friday, 6 October 2017 at 09:15 for the degree of Doctor ofPhilosophy (Faculty of Pharmacy). The examination will be conducted in English. Facultyexaminer: Assistant Research Professor Benjamin Bowen (Lawrence Berkeley NationalLaboratory, Berkeley, CA, USA).

AbstractKällback, P. 2017. Development and Application of Software Tools for Mass SpectrometryImaging. Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty ofPharmacy 234. 66 pp. Uppsala: Acta Universitatis Upsaliensis. ISBN 978-91-513-0040-5.

Mass spectrometry imaging (MSI) has been extensively used to produce qualitative maps ofdistributions of proteins, peptides, lipids, neurotransmitters, small molecule pharmaceuticalsand their metabolites directly in biological tissue sections. Moreover, during the last 10years, there has been growing demand to quantify target compounds in tissue sections ofvarious organs. This thesis focuses on development and application of a novel instrument-and manufacturer-independent MSI software suite, msIQuant, in the open access formatimzML, which has been developed specifically for quantitative analysis of MSI data.The functionality of msIQuant facilitates automatic generation of calibration curves fromseries of standards that can be used to determine concentrations of specific analytes. Inaddition, it provides many tools for image visualization, including modules enabling multipleinterpolation, low intensity transparency display, and image fusion and sharpening. Moreover,algorithms and advanced data management modules in msIQuant facilitate managementof the large datasets generated following rapid recent increases in the mass and spatialresolutions of MSI instruments, by using spectra transposition and data entropy reduction(at four selectable levels: coarse, medium, fine or superfine) before lossless compressionof the data. As described in the thesis, implementation of msIQuant has been exemplifiedin both quantitative (relative or absolute) and qualitative analyses of distributions ofneurotransmitters, endogenous substances and pharmaceutical drugs in brain tissue sections.Our laboratory have developed a molecular-specific approach for the simultaneous imagingand quantitation of multiple neurotransmitters, precursors, and metabolites, such as tyrosine,tryptamine, tyramine, phenethylamine, dopamine, 3-methoxytyramine, serotonin, gamma-aminobutyric acid (GABA), and acetylcholine, in histological tissue sections at high spatialresolution by matrix-assisted laser desorption ionization (MALDI) and desorption electrosprayionization (DESI) MSI. Chemical derivatization by charge-tagging primary amines of analytessignificantly increased the sensitivity, enabling mapping of neurotransmitters that were notpreviously detectable by MSI. The two MSI approaches have been used to directly measurechanges in neurotransmitter levels in specific brain structures in animal disease models, whichfacilitates understanding of biochemical mechanisms of drug treatments. In summary, msIQuantsoftware has proven potency (particularly in combination with the reported derivatizationtechnique) for both qualitative and quantitative analyses. Further developments will enable itsimplementation in multiple operating system platforms and use for statistical analysis.

Keywords: mass spectrometry imaging, MALDI, DESI, msIQuant, quantitation,neurotransmitters, drugs, derivatization, brain, compression, data entropy reduction

Patrik Källback, Department of Pharmaceutical Biosciences, Box 591, Uppsala University,SE-75124 Uppsala, Sweden.

© Patrik Källback 2017

ISSN 1651-6192ISBN 978-91-513-0040-5urn:nbn:se:uu:diva-328016 (http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-328016)

Acta est fabula, plaudite!

List of Papers

This thesis is based on the following papers, which are referred to in the text by the corresponding Roman numerals. For convenience, the studies reported in Papers I-V are sometimes referred to as Studies I-V, respectively.

I Källback, P., Shariatgorji, M., Nilsson, A., Andrén, P.E., Novel

mass spectrometry imaging software assisting labeled normali-zation and quantitation of drugs and neuropeptides directly in tis-sue sections. Journal of Proteomics 2012, 75, 4941–4951.

II Källback, P., Nilsson, A., Shariatgorji, M., Andrén, P.E., msI-Quant - Quantitation software for mass spectrometry imaging en-abling fast access, visualization, and analysis of large data sets. Analytical Chemistry 2016, 88, 4346–4353.

III Källback, P., Nilsson, A., Andrén, P.E., A space efficient direct

access data compression approach for mass spectrometry imag-ing. Submitted 2017.

IV Shariatgorji, M., Nilsson, A., Goodwin, R.J.A., Källback, P., Schintu, N., Zhang, X., Crossman, A.R., Bezard, E., Svennings-son, P., Andrén, P.E., Direct targeted quantitative molecular im-aging of neurotransmitters in brain tissue sections. Neuron 2014, 84, 697–707.

V Shariatgorji, M., Strittmatter, N., Nilsson, A., Källback, P., Al-varsson, A., Zhang, X., Vallianatou, T., Svenningsson, P., Good-win, R.J.A., Andrén, P.E., Simultaneous imaging of multiple neurotransmitters and neuroactive substances in brain by desorp-tion electrospray ionization mass spectrometry. NeuroImage 2016, 136, 129–138.

Reprints were made with permission from the publishers.

List of Additional Papers

Shariatgorji, M., Källback, P., Gustavsson, L., Schintu, N., Svenningsson, P., Goodwin, R.J.A., Andrén, P.E., Controlled-pH tissue cleanup protocol for sig-nal enhancement of small molecule drugs analyzed by MALDI-MS imaging. Analytical Chemistry 2012, 84, 4603–4607.

Shariatgorji, M., Nilsson, A., Källback, P., Karlsson, O., Zhang, X., Sven-ningsson, P., Andrén, P.E., Pyrylium salts as reactive matrices for MALDI-MS imaging of biologically active primary amines. Journal of The American Society for Mass Spectrometry 2015, 26, 934-939.

Shariatgorji, M., Nilsson, A., Bonta, M., Gan, J., Marklund, N., Clausen, F., Källback, P., Lodén, H., Limbeck, A., Andrén, P.E., Direct imaging of ele-mental distributions in tissue sections by laser ablation mass spectrometry. Methods 2016, 104, 86-92.

Contents

Introduction ................................................................................................... 11 Mass Spectrometry Imaging Instrumentation – Ion Sources ................... 12 

Matrix-Assisted Laser Desorption Ionization (MALDI) ..................... 12 Matrix Selection and Application ........................................................ 12 Desorption Electrospray Ionization (DESI) ......................................... 14 Other Ion Sources ................................................................................ 15 

Mass Spectrometry Imaging Instrumentation – Mass Analyzers ............. 16 Time-of-Flight (TOF) .......................................................................... 16 Quadrupole Time-of-Flight (Q-TOF) .................................................. 16 Fourier Transform Ion Cyclotron Resonance (FTICR) ....................... 17 Other Mass Analyzers Used in Combination with MSI ...................... 17 

Mass Spectrometry Imaging – Data Processing and Software ................. 18 MSI Data Processing ........................................................................... 18 Software Used for Analyzing MSI Data .............................................. 18 

Mass Spectrometry Imaging – Quantitation ............................................. 20 Absolute Quantitation .......................................................................... 20 Quantitation of Neurotransmitters ....................................................... 22 

The Development of msIQuant ................................................................ 23 

Aims .............................................................................................................. 24 

Material and Methods ................................................................................... 25 Animals and Treatments ........................................................................... 25 Tissue Section Preparation ....................................................................... 26 

Drug Tissue Section Preparation ......................................................... 26 Neurotransmitter Tissue Section Preparation ...................................... 26 

Mass Spectrometry Imaging ..................................................................... 26 msIQuant – Instrument- and Manufacturer-Independent MSI Software Using the Standardized Open Access Data Format imzML ..................... 27 

Use of the imzML Format ................................................................... 27 Implementation of MSI Data Processing ............................................. 27 Implementation of Spectra Transposition ............................................ 28 Creation and Types of ROIs ................................................................ 28 The Quantitation Process with msIQuant ............................................ 29 Data Evaluation and Export ................................................................. 29 

Data Entropy Reduction and Compression of Spectra Data .................... 30 Implementation of Relative and Absolute Quantitation of Neurotransmitters in Brain Tissue Sections by MALDI .......................... 30 Implementation of Relative and Absolute Quantitation of Neurotransmitters and Neuroactive Substances in Brain Tissue Sections by DESI ..................................................................................... 31 

Results and Discussion ................................................................................. 32 msIQuant Software ................................................................................... 32 

Data Handling in the imzML Format .................................................. 36 MSI Data Processing ........................................................................... 37 Spectra Transposition .......................................................................... 39 Evaluation Process in msIQuant – from ROI Creation to Evaluation Data Export .......................................................................................... 40 

Data Entropy Reduction and Compression of Spectra Data .................... 42 Relative and Absolute Quantitation of Neurotransmitters in Brain Tissue Sections by MALDI ...................................................................... 46 Relative and Absolute Quantitation of Neurotransmitters and Neuroactive Substances in Brain Tissue Sections by DESI ..................... 49 

Conclusions ................................................................................................... 52 

Future Developments .................................................................................... 54 

Populärvetenskaplig sammanfattning ........................................................... 55 

Acknowledgements ....................................................................................... 57 

References ..................................................................................................... 59 

Abbreviations

3-MT 3-Methoxytyramine 5-HT Serotonin (5-hydroxytryptamine) 9AA 9-aminoacridine A24 Cingulate cortex (area 24) aca Anterior commissure, anterior part Acb Accumbens nucleus AcbSh Accumbens nucleus, shell region ACh Acetylcholine ACN Acetonitrile AD Alzheimer’s disease alpha-GPC L-alpha-glycerylphosphorylcholine CA1 Cornu ammonis (field 1) CHCA α-Cyano-4-hydroxycinnamic acid CPu Caudate putamen DA Dopamine Da Dalton DESI Desorption electrospray ionization DG-sg Granule cell layer of the dentate gyrus DHB 2,5-dihydroxybenzoic acid DOPAC 3,4-dihydroxyphenylacetic acid DPP-TFB 2,4-Diphenyl-pyrylium tetrafluoroborate DPPC dipalmitoylphosphatidylcholine DT Dithranol FID Free induction decay FTICR Fourier transform ion cyclotron resonance GABA Gamma-aminobutyric acid Glu Glutamate GUI Graphical user interface HPA 3-Hydroxypicolinic acid IHC Immunohistochemical IHS Intensity-hue-saturation i.p. Intraperitoneal i.v. Intravenous ITO Indium tin oxide LZ77 Lempel-Ziv 1977 compression algorithm LZMA Lempel-Ziv-Markov chain algorithm

MALDI Matrix-assisted laser desorption ionization MeOH Methanol MDI Multiple document interface MFC Microsoft foundation class library MPTP 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine MS Mass spectrometry MS Medial septal nucleus MSI Mass spectrometry imaging MS/MS m/z

Tandem mass spectrometry Mass-to-charge ratio

NAS Network attached storage Nd:YAG Niobium doped yttrium aluminum garnet PCA Principal component analysis PD Parkinson’s disease PET Positron emission tomography Pir Piriform cortex PLS Partial least squares PNG Portable network graphics p.o. Per os Q1 First quartile Q3 Third quartile Q-TOF Quadrupole time-of-flight RF Radio frequency RMS Root mean square ROC Receiver operating characteristic ROI Region of interest SA Sinapinic acid SMST Sorted mass spectrum transform SMQ1 Simple moving first quartile SP Substance P sp Pyramidal layer of CA1 TEA Triethylamine TFA Trifluoroacetic acid THAP Trihydroxyacetophenone TIC Total ion count TIF Tagged image file format TMP-TFB 2,4,6-Trimethyl-pyrylium tetrafluoroborate TOF Time-of-flight TOF/TOF Tandem time-of-flight VDB Vertical limb of the diagonal band

11

Introduction

Mass spectrometry imaging (MSI) is a powerful technique for studying spatial distributions of biomolecules1 such as lipids,2-5 proteins,1,6 peptides,7 endoge-nous metabolites and neurotransmitters,8-10 drugs and their metabolites11-13 di-rectly in tissue sections. This is done through acquisition of mass spectra from raster points in the sections that provide information about all the biomole-cules present at detectable concentrations (which depend on the specific bio-molecules’ chemical nature, the chemical environment, as well as instrumen-tal sensitivity and settings). The most commonly used MSI systems for stud-ying biomolecules generate spectra via matrix-assisted laser desorption ioni-zation (MALDI) or desorption electrospray ionization (DESI). Regardless of the equipment used, MSI is an ex vivo technique, where the analyzed samples are obtained from an investigated organism, and some sample preparation is essential. In all MSI methods, biological tissue, usually frozen but sometimes embedded in a suitable medium,14-16 is cut with a microtome and the thin tissue sections are thaw-mounted onto a target that may be a microscope glass slide. For MALDI-MSI, a thin layer of matrix (a small organic compound with suit-able properties) must be added to the sections to facilitate the desorption and ionization of biomolecules when the tissue is radiated by a pulsating laser. In contrast, for DESI-MSI analyses simply thaw-mounting sections on glass slides is sufficient.

The most frequently used instruments for detecting ions generated from samples in MSI are time-of-flight (TOF), quadrupole time-of-flight (Q-TOF), Fourier transform ion cyclotron resonance (FTICR) and ion trap (e.g., or-bitrap) mass spectrometers.

In order to interpret MSI data, powerful software (freeware or commercial) is needed that can process the data and visualize and evaluate the results. MSI has mostly been used for qualitative studies to date,17-20 but just over ten years ago advances enabled its use in quantitative (relative or absolute) studies.21-31 In addition, recent improvements in both spatial and mass resolution have dra-matically increased amounts of MSI data produced.18,22,32,33 This has increased the need for efficient, lossless MSI data compression.34,35

MALDI- and DESI-MSI has proven especially useful for analysis of small molecules such as pharmaceutical drugs and endogenous neurotransmitters and metabolites. They enable researchers to visualize effects of new drug ther-apies by determining precise distributions of drugs and drug metabolites in a

12

tissue section. They can also provide complementary information to that ob-tained via other analytical methods and imaging techniques used in drug dis-covery and development studies. The speed, sensitivity, and molecular speci-ficity of modern mass spectrometers enable the direct simultaneous imaging of molecules in tissue sections and specific tissue microstructures at near-cel-lular spatial resolutions. However, some biomolecules of interest such as neu-rotransmitters are notoriously difficult to ionize by MALDI. Thus, an im-portant challenge for MSI researchers is to develop effective methods for chemically derivatizing these neurotransmitters,8,10 i.e., chemically modifying them to improve their ionization. This would greatly facilitate (inter alia) ex-ploration of mechanisms involved in neurological diseases such as Parkin-son’s disease8,10 where neurotransmitters play important roles.

This thesis describes the development and proof-of-principle implementa-tion of msIQuant software,22 which was primarily intended to enable absolute quantitation of target analytes from MSI datasets, but can also be used in qual-itative studies. In addition, msIQuant offers well-developed visual features, incorporating many types of pixel interpolation and functions such as image sharpening. The thesis also describes and illustrates the utility of a new MSI data compression method.

Mass Spectrometry Imaging Instrumentation – Ion Sources Matrix-Assisted Laser Desorption Ionization (MALDI) MALDI mass spectrometry (MS)36-38 has proven utility as a versatile tech-nique for assaying biomolecules. Several types of lasers have been used in MALDI systems for desorption and “soft ionization” (i.e., ionization resulting in little fragmentation of analytes). Nitrogen gas lasers are important UV laser sources (λ=337 nm), which provide good ionization efficiency. However, they have limitations such as a maximum pulse frequency of 50 Hz and relatively short service life, 108 pulses.39 Thus, solid state laser, niobium doped yttrium aluminum garnet (Nd:YAG) lasers have largely replaced them. A Nd:YAG laser (λ=355 nm) has a more complex structure but a longer life span, 1011 pulses,40 and higher maximum pulse rate of 10 kHz.40

Matrix Selection and Application Before tissue sections are exposed to a laser in MALDI-MS, a matrix consist-ing of a small organic compound must be added to assist the laser-mediated ionization, ablation and desorption of biomolecules37,38 (Figure 1). However, all matrices have distinct properties and suitability for specific classes of ana-lytes,41-46 as illustrated in Table 1.

13

Table 1. Common MALDI matrices for indicated classes of analytes.

Matrix Analytes

Sinapinic acid (SA) Proteins and peptides (>5000 Da)

2,5-dihydroxybenzoic acid (DHB) Peptides (<5000 Da), nucleotides, small molecules, lipids

α-cyano-4-hydroxycinnamic acid (CHCA) Small molecules, peptides, and proteins

Trihydroxyacetophenone (THAP) Carbohydrates

Dithranol (DT) Lipids

3-Hydroxypicolinic acid (HPA) Nucleotides, oligonucleotides

9-aminoacridine (9AA) Various metabolites

For example, 2,5-dihydroxybenzoic acid (DHB) is a highly effective matrix for MALDI-MSI of peptides, while sinapinic acid (SA) is better for proteins. The selected matrix is dissolved at a high concentration in a volatile solvent, then applied to tissue sections, generally in the last step in sample preparation for MALDI-MSI. It is important to apply the matrix homogeneously to each tissue section sample47-51 without causing any delocalization of the analytes. The optimal method of matrix application depends on the required spatial res-olution and abundance of the analyte(s) of interest, and can be performed man-ually or automatically.47-52 Manual spray deposition, using an airbrush or neb-ulizer, is a practical and economical method but the results are variable and highly dependent on the expertise of the operator. Thus, automated spotting or spraying matrix application systems are generally used. Solvent-free or dry matrix deposition methods, involving sublimation of the matrix in vacuum, have also been developed.53-55

14

Figure 1. Schematic diagram of a MALDI ion source.37,38 The pulsed laser beam ini-tiates desorption/ionization processes in the matrix and co-crystalized analytes. Ions are guided towards the acceleration grids of the mass spectrometer inlet.

Desorption Electrospray Ionization (DESI) DESI is a type of ionization in which analytes in a sample are electrically charged, under ambient conditions, by an “electrospray” of solvent molecules (that are electrically charged, positively or negatively) via a high voltage cou-pling.56 The solvent is discharged by an emitter capillary surrounded by a co-axial capillary emitting a nebulizing gas (usually nitrogen) to prevent droplet formation. Ions from the electrospray transfer their electrical charge to bio-molecules and substances in the sample to be desorbed and transferred to the mass spectrometer via an extended inlet capillary. The tissue sample, usually mounted on a microscope glass, is placed on an XYZ table. Movement in XY directions creates a raster image during spectra collection, and the spatial res-olution of DESI can be adjusted down to approx. 50 μm.57 Movement in the

15

Z-axis is used to adjust the sample’s position for optimal setting of the elec-trospray capillary and mass spectrometer’s inlet capillary (Figure 2). The ma-jor advantage of DESI is that no sample preparation is required to analyze the sample,56 unless specific biomolecules and substances are to be analyzed that are not sufficiently ionized without preparation.8 Another major advantage of the DESI method is that it can be used, in principle, with any ESI mass spec-trometer.

Figure 2. Schematic diagram of a DESI ion source during an MSI analysis (Figure from Paper V). An organic solvent-water mixture is passed through a fused silica ca-pillary with a high voltage connection at a flowrate of 1 – 3 µl/min. Dry nitrogen gas is passed through an outer nebulizer capillary at a velocity close to the speed of sound.56 In combination with the nebulizer, the electrospray desorbs and ionizes molecules in the sample, which are introduced to the inlet capillary of the mass de-tector by electrostatic force.

Other Ion Sources Secondary ion mass spectrometry (SIMS) In SIMS,58,59 a focused beam of primary ions, e.g., Cs+,60 Bi+,61 or C60

+,62,63 with energies between 103 and 105 eV sputters the surface of a sample.64,65 Secondary ions of analytes in the sample are formed, desorbed and introduced to a mass analyzer, of sector,66,67 quadrupole,68 or time-of-flight69 type. Bio-molecules that are analyzed by SIMS are usually lipids65 and elemental metal ions.70 The advantages of SIMS are high sensitivity and spatial resolution (<100 nm), which allows sub-cellular analyses.33 The disadvantage is limited mass range, up to 1,500 Da.71

16

Nanospray desorption electrospray ionization (nano-DESI) Like DESI, in nano-DESI72 ions are generated under ambient conditions. However, liquid extraction is coupled to high voltage ionization of the liquid to ionize and desorb biomolecules.72 This can substantially improve the ex-traction and ionization of biomolecules, but optimizing the fluid coupling is complex, and the spatial resolution can currently reach approx. 20 μm.73 The raster movements required to create an image are provided by an XY table that moves the sample relative to the fluid coupling.

Mass Spectrometry Imaging Instrumentation – Mass Analyzers The most frequently used mass analyzers for MSI are time-of-flight (TOF), quadrupole time-of-flight (Q-TOF), ion trap (e.g., orbitrap), and Fourier trans-form ion cyclotron resonance (FTICR) instruments.

Time-of-Flight (TOF) A TOF MS is a mass analyzer that determines the mass-to-charge ratio (m/z) of an ion by accelerating it in an electric field and then letting it travel a certain distance in a field-free drift region.74 The ion is then registered by a detector and the drift time is measured. Since all ions are accelerated with the same kinetic energy, their m/z ratio can be easily calculated as it is a function of the square of the drift time; an ion with high m/z travels more slowly than an ion with low m/z and thus has a longer drift time.

The main advantage of TOF instruments is that they can provide very large frequencies of mass determinations per unit time, up to 10 kHz, enabling very high sampling speeds, up to 50 pixels·s-1.75 In contrast to TOF, the other listed types of mass analyzers collect ions in various kinds of ion traps before anal-ysis, which decreases the maximum detection frequency. The disadvantage of TOF instruments is that recorded mass shifts may occur if there are variations in the height of a surface analyzed by MALDI, and thus variations in the drift times of given ions. Nevertheless, the main MSI applications for TOF are in MALDI-MSI, and recent introduction of MALDI TOF-TOF technology has enabled synergistic exploitation of the advantages of both MALDI ionization and tandem TOF-MS, i.e., tandem MS (MS/MS).

Quadrupole Time-of-Flight (Q-TOF) A Q-TOF mass analyzer is a hybrid instrument that differs from traditional TOF systems by incorporation of a series of quadrupoles, hexapoles or octupoles that act as ion transmitters, mass filters and collision cells before the ions enter the

17

TOF unit for accurate mass determination.76 The advantage of Q-TOF relative to TOF/TOF is that the precursor ions and fragment ions are handled more ac-curately, because they are accelerated in the TOF stage at a precise time.76 To compensate for variations due to factors such as temperature drift, a lock mass (a defined substance with known mass and fragment pattern) can be introduced and used to calibrate the system to further enhance mass accuracy.76 Moreover, the variability in m/z ratios induced by variation in surface elevation is elimi-nated in Q-TOF systems because measurement of drift time starts after the ions are injected into the TOF unit. In some Q-TOF instruments, there is an ion optic unit, which physically separates isobaric ions by their mobility. Q-TOF is ap-plied in both MALDI and DESI MSI analyses.

Fourier Transform Ion Cyclotron Resonance (FTICR) In an FTICR mass analyzer, ions are captured in a “Penning trap”, consisting of a constant magnetic field and a radio frequency (RF) field that induces the ions to rotate at their cyclotron frequency.77-79 When the RF field stops, the ions continue to rotate at this frequency and emit electromagnetic radiation that detectors capture. The process is called free induction decay (FID), and the signal is transformed from its time domain to frequency domain through Fourier transform to form a mass spectrum. The big advantages of FTICR are high mass resolving power (up to 10,000,000; defined as m divided by the peak width required for separation at that mass) and the ability to separate isobaric ions. The downside is the time required for the mass analysis, which increases with increases in mass resolution. The main MSI application is in MALDI-MSI.

Other Mass Analyzers Used in Combination with MSI Third party manufacturers of MALDI and DESI ion sources have adapted their products for use as front ends to various mass spectrometers. However, the types of mass analyzers most commonly used with these ion sources are ion trap systems, such as orbitraps,80-83 in which ions are initially introduced into a curved linear (“C”) trap. They are then passed to an orbitrap, consisting of an inner spindle-like electrode where ions are trapped, orbit and oscillate axially around the electrode due to electrostatic attraction. The detector, a bar-rel-shaped outer shell, detects the axial oscillatory motion of the ions. Mass spectra are formed by transforming the signal from the detector from the time domain to the frequency domain via Fourier transform. An orbitrap mass an-alyzer has a resolving power of up to 500,000. The advantage of orbitrap com-pared to FTICR is that the detector does not require a strong constant magnetic field and associated cryostat. The downside is that it has lower maximum re-solving power than FTICR systems, which limits separation of nearly isobaric ions.

18

Mass Spectrometry Imaging – Data Processing and Software

MSI Data Processing Raw spectra obtained from MSI experiments can rarely be evaluated without within- and/or between-spectrum data processing.28,84 Within-spectrum data processing steps typically include baseline correction, baseline subtraction, spectrum smoothing, and spectrum recalibration, while between-spectrum processing is required for comparison of spectra and involves spectra normal-ization. The latter is required to account for variations among spectra due to artifacts arising from factors such as heterogeneous matrix distribution, vari-ations in tissue structures caused by pH and salt gradients,85-87 variations in lipid background,87 and/or gradual attenuation of laser intensity due to con-tamination.87 The most common normalization methods are total ion current (TIC), median87 and root mean square (RMS) normalization.87 In addition, an analyte can be normalized in relation to an appropriate internal standard, if present.22-24,26,28,30 If mass shifts occur and spectrum recalibration is insuffi-cient, unsupervised spectra correction22,88-93 might be needed.

Software Used for Analyzing MSI Data One of the first software packages developed for analyzing MSI data was the freeware Biomap.94 Some years later, commercial software packages were de-veloped by MS instrument manufacturers, such as flexImaging (Bruker Dal-tonics, Bremen, Germany) and ImageQuest (Thermo Scientific, San Jose, CA, USA). The MSI software at this time could only generate qualitative data in the form of imaging ion distributions, and instrument data could only be ex-ported in Analyze Image 7.5 format.95 Therefore, many research groups de-veloped in-house scripts to display information drawn from raw data they ac-quired.9,28,96,97

The first software developed by our research group was a suite designed to extract, process, normalize and quantify data generated in the raw data format XMAS (Bruker Daltonics), which was published as downloadable freeware.28 Development of MSI software was facilitated by the release in 2011 of the open MSI data format imzML,35 which is supported by most instrument man-ufacturers today, and various freeware/open access and commercial software packages are now available (see Table 2).

The commercial MSI software packages available are flexImaging (Bruker Daltonics), HD Imaging (Waters Corp., Manchester, UK), TissueView (AB Sciex, Concord, ON, Canada), MALDIVision (Premier Biosoft, Palo Alto, CA, USA), Quantinetix (Imabiotech, Loos, France) and SCiLS Lab (SCiLS,

19

Bremen, Germany). HD Imaging has been developed for imaging and visual-ization of proprietary instrument data with ion mobility. TissueView is a fur-ther development of Biomap, MALDIVision is a qualitative analysis software package, while Quantinetix is intended for quantitative analysis, and SCiLS Lab for different statistical analyses.

The freeware/open access MSI software packages available are Spec-tViewer,98 DataCube Explorer,99 Mirion,100 MSiReader,101 Omnispect,102 OpenMSI,34 Cardinal,103 msIQuant22 and SpectralAnalysis.104 All these pack-ages enable qualitative analysis, but Cardinal and SpectralAnalysis also offer statistical analysis capacities, and msIQuant has been specifically developed for quantitative analyses.

Table 2. Commercial and freeware/open access MSI software packages.

Software Organization/Vendor Freeware / open access

Release Year

Biomap Novartis Yes 2002

flexImaging Bruker Daltonics No 2005

ImageQuest Thermo Fisher Scientific No 2007

TissueView AB Sciex No 2010

HD Imaging Waters Corp. No 2011

MALDIVision Premier Biosoft No 2012

Quantinetix Imabiotech No 2012

SpectViewer CEA (Commissariat à l’énergie atomique) Yes 2013

DataCube Explorer FOM Institute AMOLF Yes 2013

Mirion Justus-Liebig University Giessen Yes 2013

MSiReader North Carolina State University Yes 2013

Omnispect The Georgia Institute of Technology Yes 2013

OpenMSI Lawrence Berkeley Na-tional Lab Yes 2013

SCiLS Lab SCiLS No 2013

Cardinal Northeastern University Yes 2015

msIQuant Uppsala University Yes 2016

SpectralAnalysis Birmingham University Yes 2016

20

Mass Spectrometry Imaging – Quantitation Absolute Quantitation MALDI-MSI was initially used to detect large biomolecules, such as proteins, peptides and lipids.1-6 In the lower mass range of drug molecules and metabo-lites, there may be strong interference from matrix molecules yielding isobaric ions that are difficult to separate from analytes of interest by mass analyzers with insufficient resolving power.105 Due to artefactual effects of sample prep-aration on spectra that contribute to image noise, such as variations in matrix application and degrees of co-crystallization, quantitative MALDI-MSI has previously been difficult. However, many research groups have demonstrated that MALDI-MSI can be used quantitatively,10,12,21-24,26,28,30,106,107 and many obstacles can be overcome by using new matrices,10,108 methods for applying matrices,47,48,50,53-55 and/or internal standards that minimize effects of arti-facts22-24,26,28,30 (see Figure 3). In addition, to increase selectivity and circum-vent problems caused by isobaric ions, tandem MS can be used.12,24,107

Now, MALDI-MSI analysis and high-resolution histology enables direct quantitation of in vivo drug uptake in targeted tissue structures.12 For example, my colleagues and I (hereafter we) have presented a MALDI-MSI-based method for quantifying and tracking in vivo transport of drugs directly in sec-tions from specific organs and tissues.12 We have also demonstrated the im-portance of appropriate normalization, by comparing calibration curves for imipramine obtained using different normalization procedures, which differed substantially in linearity and the dramatic improvement of the coefficient of determination (R2).28 The use of deuterated internal standards significantly im-proves the normalization of target analytes in tissue sections and increases pixel-to-pixel precision.22,24

Effects of applying standards and matrices in various ways have been ex-amined in efforts to enhance quantitative results. For example, calibration standards have been applied under tissue, onto tissue or directly on MALDI targets. Similarly, internal standards have been applied under tissue, onto tis-sue or together with matrix.22-24,26,28,30 Generally, the best results seem to be achieved when calibration standards are applied onto tissue sections, internal standard is evenly distributed on the tissue, and matrix application is the last step.22,23 To confirm that absolute MSI quantitation is possible, the protocol must be cross-validated with another established quantification method, such as liquid chromatography (LC)-MS/MS.12,24,107 Absolute quantitation proto-cols based on use of DESI MSI8 and nano-DESI MSI9 have also been reported.

21

Figure 3. Schematic illustration of workflow in a quantitative MALDI-MSI analysis with internal standard application, following administration of citalopram (30 mg/kg) to mice by intraperitoneal (i.p.) injection. A dissected mouse brain, snap-fro-zen in dry ice-cooled isopentane (A) is subjected to coronal sectioning in a cry-omicrotome (B). Brain tissue sections are thaw-mounted onto indium tin oxide (ITO) glass slides (MALDI targets) (C), then analyte calibration standards are spot-ted on control tissue (D). Internal standard and matrix are evenly applied by an auto-matic spraying device (E and F, respectively). A laser (red) ablates the matrix and desorbs/ionizes analytes in the sample (black streaks) (G). Spectra are acquired pixel-by-pixel (H). Finally, TIC-normalized analyte (citalopram [M+H]+ m/z 325), internal standard (citalopram-D6 [M+H]+ m/z 331), and analyte normalized against internal standard (325/331 m/z signal ratio) images are generated (I).

22

Quantitation of Neurotransmitters Neurotransmitters are small molecules, e.g., the catecholamine dopamine (DA), amino acids such as γ-aminobutyric acid (GABA) and glutamate (Glu), and acetylcholine (ACh). They are important messengers in the brain, which transmit signals between the neurons. Changes in their concentrations are as-sociated with many normal neural processes, such as sleep and aging, but also many diseases such as Alzheimer’s disease, Parkinson’s disease (PD) and de-pression. Thus, better understanding of their relative abundance and distribu-tions would provide insights into these complex neurological processes and disorders.

Until recently, researchers have relied on indirect histochemical, immuno-histochemical (IHC), and ligand-based methods to detect these small molecule neurotransmitter substances.109-111 However, antibodies used in IHC have many limitations, such as lack of specificity and binding to different types of transmitters.112 Nuclear medicine imaging technologies, such as positron emission tomography (PET) and single photon emission computed tomogra-phy,113 are widely used for indirect visualization of the activity, abundance, and distributions of neurotransmitters in the brain. However, developing ap-propriate radiolabeled substances is often complicated, especially for studying endogenous chemical messengers.114 In addition, it is often impossible to dis-tinguish between signals from labeled target molecules and their metabolites with residual labeling.

Thus, there is a clear need for imaging techniques that can directly visualize distributions of neurotransmitters and quantify them without labeling. MALDI-MSI is an analytical technique that offers the ability to detect and quantify neurotransmitters directly in brain tissue sections, but until recently its potential has been limited by their poor ionization efficiency.10 However, we have demonstrated that neurotransmitters with primary amines can be chemically derivatized by the pyrylium cations 2,4-diphenyl-pyrylium tetra-fluoroborate (DPP-TFB) or 2,4,6-trimethyl-pyrylium tetrafluoroborate (TMP-TFB) to form positively charged quaternary pyridinium ions,115,116 which ab-sorb light at wavelengths of the Nd: YAG laser (λ=355 nm) and thus provide a reactive matrix. Neurotransmitters and endogenous substances that can be measured following this derivatization include tyrosine, tryptamine, phene-thylamine, DA, 3-methoxytyramine (3-MT), serotonin (5-HT), GABA and Glu. Furthermore, we have been able to image and quantify acetylcholine (ACh) and the ACh precursor (L-alpha-glycerylphosphorylcholine (alpha-GPC) in brain tissue sections by using CHCA-D4 as a matrix. This form of deuterated CHCA was used because the mono isotopic m/z value of ACh (m/z 146.118) is overlapped by the peak from CHCA [M-CO2+H]+ (m/z 146.060). In addition, the DESI MSI method can ionize derivatized neurotransmitters directly in tissue sections, for imaging and quantification.8

23

The Development of msIQuant Most software packages for visualizing MSI data are solely designed for qual-itative analysis. Furthermore, the software tools for acquiring and visualizing MSI data are usually instrument vendor-specific and proprietary. Therefore, in the work this thesis is based upon we developed a novel MSI software pack-age that includes protocols for quantifying drugs and endogenous compounds (Paper I). The program can be used to process signals from predefined re-gions-of-interest (ROI) in a tissue section, retrieve corresponding mass spec-trometric data, and apply several spectral processing techniques, including baseline correction, baseline subtraction and denoising, spectrum smoothing, spectrum recalibration, and normalization. However, this software package had limitations that prohibited its application by a broader community, which prompted improvements detailed in Paper II. The following msIQuant software requirements and functionality were specified:

Independence of instrument manufacturer, through use of the input

data format imzML.35 Introduction of a spectral transposition function22,34 for fast data ac-

cess and ability to accommodate data on network attached storage (NAS) and analyze data stored in them.

Appropriate data structure to describe entire MSI experiments and ac-

commodate all generated data. A graphical user interface with a fast flicker-free graphic engine for

quick image update when changing and/or selecting (for example) one or several m/z values.

The possibility to create ROIs with different shapes.

The ability to create automatic standard curves and display the quan-

titation results. The ability to extract information and graphic images from all dialogs

and windows.

With these specifications, the software would be applicable for all types of MSI experiments, regardless of the type of ion source and mass analyzer used to generate the MSI data.

24

Aims

This thesis is based upon a series of studies in which algorithms and software for qualitative and quantitative evaluation of MSI data were developed and implemented. The overall purpose was to develop a powerful software pack-age for accessing and evaluating the very large datasets generated when quan-tifying drugs and endogenous compounds in tissue areas of interest, and for processing associated mass spectra and images. Specific aims of the studies, reported in Papers I-V, were to:

Investigate and develop essential software procedures for analyzing MSI data, such as spectra processing, normalization, quantitation, and data evaluation (Paper I).

Investigate and develop fully-featured software for MSI data evalua-

tion and quantification, regardless of the manufacturer of the instru-ments used and MSI data format (Paper II).

Develop space-efficient direct data access and data compression pro-

tocols, then investigate and exemplify results provided by the meth-odology (Paper III).

Utilize the prototype software for quantitative measurements of neu-

rotransmitters in brain tissue sections by MALDI-MSI (Paper IV).

Utilize the developed msIQuant software to quantify neurotransmit-ters and neuroactive substances in brain tissue sections probed by DESI-MSI (Paper V).

25

Material and Methods

Animals and Treatments All rodents used in the studies were housed in air-conditioned rooms with a 12 h dark/light cycle at 20 °C, 53% relative humidity and food ad libitum. The studies were conducted in accordance with European Community Council Di-rectives of 1986 (86-609-EEC) and 2010 (2010/63/EU), and were approved by the local ethical Committees for animal experiments. Adult male rats (Wistar or Sprague-Dawley) were used in Studies I – V. In Studies I and II, Wistar rats were administered tiotropium bromide by in vivo inhalation or im-ipramine per os (p.o.). In Studies III-V, Sprague-Dawley rats were unilaterally lesioned by injecting 6-OHDA into the medial forebrain bundle (MFB) of the right hemisphere. Four weeks after the surgery, the animals were treated with L-DOPA and benserazide interperitoneal (i.p.) injection once daily for 3 weeks. Animals were sacrificed by decapitation 30 min after the last i.p. injection. Male mice (C57BL/6), adult and 3 months old, were used in Studies I – V. In Study I, mice were administered imipramine by i.p. injection. In Studies II and III, mice were administered saline by i.p. injection. In Study IV, mice were administered saline or tacrine by i.p. injection. In Study V, mice were admin-istered fluvoxamine, sibutramine, or amphetamine, by i.p. injection. Follow-ing euthanasia, brain or lung tissue was rapidly dissected, frozen in dry ice-cooled isopentane, and stored at -80 °C until further use. Primate experiments were conducted using tissue from a previously published brain bank 117-119. Experiments were carried out in accordance with European Community Council Directive of November 24, 1986 (86/609/EEC) for care of laboratory animals. Brain tissue from two female, adult (5±1 years) rhesus monkeys (Macaca mulatta) was used in Study IV. One rhesus monkey was untreated (control) while the other was administered with 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) by intravenous (i.v.) injection according to a previously published protocol.120-122 The brain sections were stored at -80 °C until preparation and analysis.

26

Tissue Section Preparation Brain and lung tissue sections, 12 or 14 μm thick, were cut using a cryomicro-tome (Leica CM3050S, Leica Microsystems, Germany) at a temperature of -20 °C and thaw-mounted onto conductive indium tin oxide-coated (ITO) slides (Bruker Daltonics) or conventional microscope glass slides. The sec-tions were stored at -80 °C until required. On the day of analysis, the tissue sections were desiccated at room temperature for 20 min and optical images were acquired using a standard flatbed scanner (Seiko Epson, Japan) prior to further preparation.

Drug Tissue Section Preparation Serial dilutions of the target compounds, tiotropium, imipramine and sub-stance P (SP), were manually spotted onto corresponding control tissue sec-tions. Internal standard (IS) was either mixed with MALDI matrix or applied directly to the tissue sections using an automatic spraying device (TM-Sprayer, HTX Technologies LLC, Chapel Hill, NC). MALDI matrices were applied to the tissue sections using automatic spraying devices (ImagePrep, Bruker Daltonics and TM-Sprayer). Two matrices were used: α-cyano-4-hy-droxycinnamic acid (CHCA) or 2,5-dihydroxybenzoic acid (DHB).

Neurotransmitter Tissue Section Preparation Brain tissue sections were derivatized with DPP-TFB (Papers II - V) or TMP-TFB (Paper IV) to detect endogenous neurotransmitters or drugs with primary amines. The reactive matrix, DPP-TFB or TMP-TFB, was sprayed over the dry tissue using an automated spraying device with a spray nozzle temperature set at 80 °C. The tissue sections were then incubated for 3 × 5 min under pre-viously described conditions (Paper IV), then dried in a dry nitrogen atmos-phere.

Mass Spectrometry Imaging The MSI data presented in this thesis were acquired using ultraflex II TOF/TOF, ultrafleXtreme TOF/TOF, solariX XR 12 T (Bruker Daltonics), LTQ Orbitrap XL (Thermo Scientific) and Synapt G2si (Waters Corp.) instru-ments. In Studies II and III, a DESI (DESI 2D, Prosolia, Indianapolis, IN, USA) ion source, mounted on a Synapt G2si (Waters Corp.) instrument was used. In Study V, the MSI data were acquired using a DESI ion source (OMNIspray,

27

Prosolia, Indianapolis, IN, USA), mounted on a Q-Exactive Plus (Thermo Sci-entific) instrument. For both DESI sources, methanol/water (95:5 v/v) was used as the electrospray solvent at a flow rate of 1.5 µl/min.

msIQuant – Instrument- and Manufacturer-Independent MSI Software Using the Standardized Open Access Data Format imzML msIQuant was developed on the Microsoft Foundation Class Library (MFC) 64-bit (x64) platform, written in the C++ programming language, and the complete program consists of more than 100k lines of code. The graphical user interface (GUI) consists of a tabbed multiple document interface (MDI) and several docking windows. The main windows in the GUI are the project explorer, the spectra view, and the mass list view, which all are docking win-dows, while the image view (displaying the MSI data) is the MDI. In order to display semitransparent images, anti-aliased 2D graphics and images in vari-ous formats like TIF, jpeg and PNG, the C++ based graphics device interface GDI+ was used. To make the graphic flicker-free, a graphic display technique called double buffering was used.

Use of the imzML Format The msIQuant software has been created to read MSI data as information in imzML language,35 an open access data format for mass spectrometry imaging created in 2008 in a European academic project called COMPUTIS. Today, systems supplied by most MSI instrument vendors can generate MSI data in imzML format.

Implementation of MSI Data Processing The data processing steps that have been implemented in msIQuant are spectra processing and spectra normalization. The spectra processing options that can be performed in msIQuant are spec-trum baseline correction by a simple moving first quartile (SMQ1) algorithm, spectrum baseline subtraction and denoising by the sorted mass spectrum transform (SMST) algorithm,28 spectrum smoothing by the Savitsky-Golay al-gorithm,123 and finally spectrum recalibration. Spectrum recalibration can be performed by either automatic lock-mass free recalibration22,91 or using data pertaining to known calibrants.

28

The MSI normalization methods that are implemented in msIQuant are me-dian,87 TIC,28,84,124 RMS87 and internal standard22-24,26,28,30 normalization.

Implementation of Spectra Transposition In order to access and read the MSI data efficiently during analysis, the data in msIQuant are stored in both spectra and imaging modes.

In the imaging mode, the spectra are stored in transposed arrays (Figure 4), each registering the intensity of signals corresponding to one m/z value for every pixel in an image. This enables fast data access and storage of the data in a network attached storage (NAS) device since only a small amount of the data needs to be read.

Figure 4. Two modes of reading mass spectra using msIQuant (Figure from Paper II, supplementary information). (A) In spectrum mode, the MSI spectra are arranged spectrum-by-spectrum. (B) In imaging mode, the transposes (MT) of the MSI spectra are arranged pixel array-by-pixel array (registering the intensity of signals corre-sponding to one m/z value for every pixel in an image). This enables rapid access of the m/z data. The arrows indicate the reading direction in the computer memory.

Creation and Types of ROIs Three types of ROIs can be created in msIQuant: polygonal, rectangular, and elliptical. After ROI creation, the ROIs can be annotated, e.g., to refer to a specific structure of the tissue. Elliptical ROIs are often useful when adding ROIs of standard spots, while polygonal ROIs are often needed when adding ROIs of certain tissue structures.

29

The Quantitation Process with msIQuant The quantitation process using msIQuant is based on adding standard calibra-tion spots to control tissue. A new quantitation entry is created by generating ROIs of the standards then selecting the source for the quantitation (m/z value or labeled normalization), as illustrated in Figure 5.

Once the quantitation entry is created, the amount of the corresponding substance in each pixel is calculated and presented in the quantitation evalua-tion window/module.

Figure 5. Create Quantitation dialog window. A standard curve is created automati-cally when the absolute amount of a standard at each spot is added. After accepting the standard curve, the quantitation is expressed as the amount of substance per pixel in the experiment.

Data Evaluation and Export Data can be evaluated when the source for visualization (m/z value, internal standard normalization, or quantitation), normalization method, and ROIs have been created and annotated. The following basic statistics are calculated for each ROI: number of measurement pixels, area, perimeter, volume, mass, sum of pixel intensities, average intensity, standard deviation, median inten-sity, Q1 intensity, Q3 intensity, minimum intensity, and maximum intensity. The evaluation list can then be copied to the clipboard and pasted into various software packages, e.g., Excel (Microsoft, Redmond, WA, USA) or MATLAB (MathWorks, Natick, MA, USA), for further evaluation.

30

In addition, all values and images, including spectra presented in msIQuant, can be copied to the clipboard and pasted into any software for further anal-yses. For publishing with high quality, the MSI images can be saved in the image formats PNG and TIF.

Data Entropy Reduction and Compression of Spectra Data A method of compressing MSI data has been developed to reduce the disc space needed to store the large datasets that often are produced in modern MSI experiments. It is based on the lossless data compression method Lempel-Ziv-Markov chain algorithm (LZMA),125 which is descended from the Lempel–Ziv 1977 algorithm (LZ77).126 To increase the compression ratio, the data en-tropy is reduced by decreasing the number of significant figures, by logarith-mic transformation and type casting the values from double precision floating points to an integer data type. The accuracy of the data entropy reduction has four settings: coarse (10-bits), medium (13-bits), fine (16-bits), and superfine (20-bits).

Implementation of Relative and Absolute Quantitation of Neurotransmitters in Brain Tissue Sections by MALDI The neurotransmitters in brain tissue sections from rat, mouse, and primate, were analyzed by a prototype version of the msIQuant software. Relative and absolute quantitation of the neurotransmitters and small molecules, e.g., do-pamine (DA), 3-methoxytyramine (3-MT), acetylcholine (ACh), L-alpha-glycerylphosphorylcholine (alpha-GPC), serotonin (5-HT), and gamma-ami-nobutyric acid (GABA) were performed. RMS normalization was applied for both the relative and absolute quantitation. All neurotransmitters and small endogenous molecules were derivatized with DPP-TFB (resulting in a 215.086 Da mass shift) except ACh and alpha-GPC, which do not have any primary amine, and instead CHCA-D4 was used as a matrix. To quantify neu-rotransmitters, isotopically labeled calibration standards were spotted on con-trol tissue to avoid any interferences originating from endogenous com-pounds. GABA-D6 standard was used for absolute quantitation of GABA. To determine absolute quantities of ACh, ACh-D9 was spotted as a standard on control tissue.

31

Implementation of Relative and Absolute Quantitation of Neurotransmitters and Neuroactive Substances in Brain Tissue Sections by DESI The msIQuant software was used to analyze amounts of neurotransmitters in rat and mouse brain tissue sections detected by DESI-MSI, and quantify (rel-atively and absolutely) amounts of two administered drugs: amphetamine and fluvoxamine. It was also used for qualitative imaging of distributions of the neurotransmitters and various small molecules, including adenosine, aspar-tate, glutamate, DA, DA-D3, 3,4-dihydroxyphenylacetic acid (DOPAC), GABA, GABA-DPP, 5-HT-DPP, DA-DPP, DA-D3-DPP, 3-MT-DPP, 4-hy-droxyamphetamine, and N-di-desmethylsibutramine. No matrix was used. In experiments including derivatization of the analytes, the derivatization agent used was DPP-TFB (which results in a 215.086 Da mass shift). Standards were spotted on control tissue to determine absolute quantities of fluvoxamine, and the internal standard fluvoxamine-D4 was applied, before derivatization, by an automatic spraying device (TM-Sprayer, HTX Technologies) for normaliza-tion. In addition, relative quantities of derivatized amphetamine were deter-mined to compare tissue concentrations in animals with different administered doses of amphetamine (1 and 5 mg·kg-1).

32

Results and Discussion

msIQuant Software The work this thesis is based upon initially involved development of a novel MALDI-MSI software package, and its application in proof-of-principle MSI quantitation, including demonstration of its utility for quantifying small mol-ecules such as drugs (Paper I). The software enables rapid performance of routine spectrum processing tasks, such as baseline subtraction, smoothing, recalibration, normalization and extraction of average spectra from regions of interest in a tissue section. It also facilitates use of an internal standard for normalization and quantitation directly in tissue sections. We demonstrated that concentrations of drugs or endogenous compounds such as neuropeptides can be semi-automatically determined using either external standard curves or isotope-labeled analogs as standards for normalization. The latter approach resulted in lower deviation between neighboring pixels and compensated for false ion signals in tissue sections.

However, the software described in Paper I had several limitations, notably it could only handle data provided by systems supplied by one manufacturer (Bruker Daltonics), operated using flexImaging software in the proprietary MSI data format XMAS (Paper I). Thus, a new, improved software package (msIQuant) for MSI quantitation was subsequently developed (Paper II) to handle data from diverse types of MSI instruments supplied by various man-ufacturers through use of the standardized open MSI data format imzML.35 msIQuant has been completely rewritten to meet specifications we defined to meet anticipated requirements for MSI experiments. New functionalities like capacities to select different types of regions of interest (ROI) and normaliza-tion methods have been implemented. However, functionality from the previ-ous software has been retained, such as spectra processing, normalization, quantitation and evaluation. The new, developed data management layer is designed to handle very large amounts of data, with annotation regarding the source MSI experiment(s). The graphical user interface (GUI) in msIQuant is utilized to access, manipulate, evaluate and export MSI data (Figure 6).

33

Figure 6. Screenshot of msIQuant’s graphical user interface (Figure from Paper II, supplementary information). The interface is divided into four panels. (A) The project panel lists experiments, which can be selected and displayed. (B) The spectra panel displays the average and the maximum intensity spectra of the image. (C) The mass list panel shows the annotated and selected ions. (D) The image panel displays the dis-tribution of the selected ion, the normalization factor and intensities/concentrations.

In addition to quantitative analysis, capacity for qualitative analysis is im-portant for purposes such as mapping ion distributions. Thus, graphic capabil-ities have been enhanced in msIQuant, enabling a broad range of visualization functions, notably the possibility to choose from four interpolation methods (nearest neighbor, bilinear, bilinear-color-blend, or bicubic) when displaying ion distributions in tissue sections. Moreover, depending on the type of anal-ysis and the user’s preferences for representing ion intensities, seven rainbow scales and four monochrome color scales are offered. The intensity can also be displayed in both linear and logarithmic scales (Figure 7).

34

Figure 7. (A-D) Illustration of the four methods for visualizing ion distributions available in msIQuant, based on Nearest Neighbor (A), Bilinear (B), Bilinear-Color-Blend (C), and Bicubic (D) interpolation. (E-K) Rainbow scales to visualize ion in-tensities: Red Temperature (E), Thermal (F), Blue-Red (G), Bruker Rainbow (H), Biomap Rainbow 1 (I), Biomap Rainbow 2 (J), and Synapt Blue-Red-Yellow (K). (L-O) Monochrome color scales: Black-White (L), Red (M), Green (N), and Blue (O). (P-Q) Ion images of citalopram [M+H]+ (m/z 325), normalized using the inter-nal standard citalopram-D6 [M+H]+ (m/z 331) displayed in linear-intensity (P) and logarithmic-intensity (Q) scales.

Three features were employed to improve ion visualization: low intensity transparency,22,94 image sharpening,22 and intensity-hue-saturation (IHS) im-age fusion.127-129 Image fusion is used when information from different mo-dalities are combined to create an image. The simplest way of combining an ion image with a registered histological image is to use different degree of transparency to overlay and view both images. There is also a way to combine and fuse the modalities statistically and predict the ion image against the fea-tures of the histological image to gain high spatial resolution.130 The image sharpening and IHS image fusion tools were implemented to combine high-resolution histological or optical images with low spatial resolution ion im-ages so detailed molecular distribution information could be correlated with specific tissue areas.

35

Figure 8. Low intensity transparency, profile intensity, and image fusion using msI-Quant (Figure from Paper II). (A) Optical image of a coronal rat brain tissue section displaying annotated brain structures: corpus callosum (cc), caudate putamen (CPu), nucleus accumbens, core region (AcbC), anterior commissure, anterior part (aca), and olfactory tubercle (Tu) at 15 μm resolution.131 The magnified inset shows the aca, AcbC, and CPu. (B) Ion image of DA (m/z 368.165, DA derivatized with DPP-TFB) with nearest neighbor interpolation at 150 μm spatial resolution. (C) The DA image was sharpened with the image fusion function of msIQuant by using the opti-cal image to display the ion distribution at a spatial resolution smaller than that of the measured ion image. (D) Ion image displaying the distribution of DA with bicu-bic interpolation at a spatial resolution of 150 μm. (E) Overlay ion image of DA with 50% opacity, with co-registered optical image. (F) The optical image is shown in the bottom layer, and the DA ion distribution with low intensity transparency is shown in the top layer. The low intensity transparency enables preservation of the brightness and contrast of both image layers. (G) The profile ion intensity of DA is selected from a defined area (200 μm wide and 6082 μm long). (H) Graph showing the relative intensity of DA (y-axis) across the long axis of the tissue section (x-axis, μm). The MS images were generated using a MALDI FTICR instrument.

The image sharpening process is exemplified by the fusion of an ion intensity image showing the distribution of dopamine at a pixel resolution of 150 μm with an optical image of the corresponding tissue section taken at 15 μm pixel resolution (Figure 8A-C). The image processing algorithm includes a protocol for detecting edges in the high-resolution histological or optical component of a fused image to highlight structures visible in the high-resolution image. The

36

implementation of low-intensity transparency enhances visualization of the intensity dynamics (Figure 8D-F). High intensity areas are completely opaque, while the transparency of areas with intensities below 15% of the maximum scale is increased. This increases the visibility of the underlying histological and/or optical image. Without the low-intensity transparency function, it is difficult to distinguish intensities close to zero. Low intensity transparency is required to achieve correct color rendering with respect to ion intensity and optical images.

In addition, a profile measurement function of the ion intensity was intro-duced that enables visualization of the intensity of a selected ion along a pre-defined line with a set width (Figure 8G-H). The intensity profile can be cop-ied and pasted into a spreadsheet, where the values along the profile can be plotted to get information about the ion intensity along the profile and its var-iation with distance from a given point.

An additional feature has been introduced to visualize ions’ distributions through red-cyan anaglyph visualization, in such a way that the histological or optical image is set at a long distance while the ion intensity is set in the foreground, and the closer to the viewer the stronger the ion intensity (this visualization feature requires red-cyan 3D glasses).

Data Handling in the imzML Format The only MSI data format that can be used to create msIQuant data is the open access MSI format, imzML. Specifically, the binary component of imzML data is converted into continuous profile data to avoid further recalculation or conversion of the data after it has been read into the memory. In imzML for-mat, data may be stored as either processed profile or centroid spectra. When converting a processed profile spectrum, all parts of the spectrum are placed in a continuous spectrum. When converting centroid spectra, each centroid is converted into a Gaussian distribution based on the mass resolution and added to a continuous spectrum.

The versatility of msIQuant is demonstrated in Paper II by displaying ion images acquired with MSI instruments from three manufacturers (Waters Corp., Thermo Scientific, and Bruker Daltonics), with different types of mass analyzer (TOF, orbitrap and FTICR). Moreover, the paper demonstrates not only the software’s supplier independence, but also its ability to use different binary formats specified in imzML (continuous profile, processed profile and processed centroid binary formats). The presented data were acquired with MS instruments equipped with DESI and MALDI ion sources, but other types of ion sources such as SIMS or nano-DESI systems could also be used since the conversion of MSI data to imzML format is independent of the ion source type.

37

MSI Data Processing Data processing of MSI spectra is necessary before evaluation of qualitative or quantitative data. The data processing can be divided into five steps: spectra processing, spectra normalization, ROI creation and annotation, evaluation, and post processing.

The most important spectra processing actions are baseline correction, baseline subtraction, denoising, smoothing, and recalibration. Depending on the post-processing procedure, baseline correction, subtraction and spectra re-calibration are the most important steps. In addition, mass shifts may occur in MSI experiments, and must be accounted for. In TOF instrumentation, they may be caused by differences in height of the analyzed samples, or variation in the mass calibration when several tissue slides are analyzed at different times. In FTICR instruments, drift in time and instrumental parameters may cause small mass shifts, which are sufficient to cause interference from near-isobaric ions. One way to overcome this problem is to use a wider mass range when selecting an m/z value. However, if the spectra are generated by FTICR systems (for example), isobaric peaks might perturb the results if a too wide mass range is used. Another, better way to solve this problem is to apply a spectra recalibration step (Figure 9, Paper II). The spectra processing steps, noise removal and spectra smoothing, also need to be carefully considered because they might impair post-processing algorithms that require noise.130

Figure 9. Automated mass alignment and recalibration (Figure from Paper II, sup-plementary information). The three colored combined peaks represent spectra of a dimer of α-cyano-4-hydroxycinnamic acid [2M+H]+ (m/z 379.092) from three tissue sections on the same glass slide. The mass spectra were acquired using a MALDI-TOF/TOF instrument (ultrafleXtreme, Bruker Daltonics). (A) Three overlaid mass spectra with three peaks at different m/z values, all corresponding to the same ion (red m/z 379.074, green m/z 379.115, and blue m/z 379.089) before mass recalibra-tion. The black peak (m/z 379.105) represents the average spectrum of the three overlaid mass spectra. (B) After mass recalibration against the average spectrum, all peaks have the same centroid mass (m/z 379.105 – mass tolerance, 34 ppm).

38

Spectra normalization is a way of increasing or decreasing a spectrum signal to minimize systematic artifacts caused by the chemistry or variation of the signal in an experiment. Such artifacts may be caused by variations in matrix crystal distribution and/or morphology of the crystals, ion source contamina-tion and laser attenuation,87 which weakens the signal over time, and ioniza-tion suppression caused (for instance) by salts, pH gradients, and phospholipid background.85-87 As already mentioned, the normalization methods that can be selected in msIQuant are median,87 TIC, RMS,87 and internal standard22-

24,26,28,30 normalization. Median, TIC, and RMS normalization affect the whole spectrum per pixel while internal standard normalization is based on the ratio between the analyte and the internal standard m/z values. Unfortunately, there are no guidelines for identifying the optimal normalization method. However, when an isotopically labeled internal standard is used, internal standard nor-malization is preferred because the analyte and the internal standard have the same chemical and physical properties, so pixel-to-pixel variation is mini-mized22,24,28 (Figure 10, Paper I).

Figure 10. Ion images, processed with different normalization methods, of imipra-mine (A-F) and tiotropium (G-L) in central lung tissue sections from rats (Figure from Paper I). (A, G) Optical images of lung tissue sections. Arrows indicate areas where no lung tissue is present. (B, H) Raw values only, and images obtained with: (C, I) median normalization, (D, J) TIC normalization, and (E, K) RMS normaliza-tion. (F, L) Standard curve obtained with internal standard (IS) normalization. The signal from imipramine [M+H]+ (m/z 281.2) was normalized by the signal from the IS, imipramine-D3 [M+H]+ (m/z 284.2), and tiotropium [M]+ (m/z 392.1) by the IS, tiotropium-D3 [M]+ (m/z 395.1).

The image quality of the ion distribution is not only the only factor to consider when selecting a normalization method, because it also strongly affects stand-ard curves generated for absolute quantitation. For example, as illustrated in Figure 11 the coefficient of determination (R2) is dramatically improved when

39

internal standard normalization is used (Paper I). However, if this normaliza-tion method cannot be applied, for example due to the lack of an analogue of the investigated analyte, the median, TIC or RMS normalization method can be used. If the raw data have been processed so that spectra noise has been removed to save storage space, only TIC or RMS normalization can be used since the median normalization factor for each spectrum, in this case is zero.

Figure 11. Calibration curves of imipramine standards obtained using different nor-malization methods (Figure from Paper I). (A) Optical image of a lung tissue section with spotted with indicated amounts of imipramine standard. (B) Standard curve based on raw values. (C) Standard curve with median normalization. (D) Standard curve with TIC normalization. (E) Standard curve with RMS normalization. (F) Standard curve with internal standard (IS) normalization. The signal from imipra-mine was normalized by the signal from the IS, imipramine-D3.

Spectra Transposition In recent years, most MSI datasets have become very large; too large to host in current workstations’ internal memory. For example, data generated in an FTICR MS experiment may require more than 500 GB of memory. In some MSI software packages this problem is resolved by binning the data,99,132,133 or cutting the beginning and/or end of each spectrum99 to reduce the number of m/z channels. Another approach that has been applied is to reduce the image area99 (i.e., number of pixels), to reduce the dataset sufficiently for storage in the memory while retaining the mass resolution. The problem of loading very large MSI datasets into computers has been re-solved in msIQuant by storing MSI data in two modes: spectrum and image

40

mode (Figure 4). Spectrum mode is the standard way of storing spectra, spec-trum-by-spectrum (one spectrum per pixel), while in imaging mode data are stored in pixel vectors (one pixel vector per m/z channel), also known as trans-posed spectra.22 Instead of creating transposed spectra, the MSI data can also be divided into a number of chunks with limited mass ranges.34 The advantage of storing data as transposed spectra, with all pixel intensities per vector, is that only one m/z array needs to be read to display the distribution of an ion with a specific m/z value, so data-read-access is very quick.

The data-read-access time depends on the number of pixels covered in a specific experiment and the data transfer speed. The amount of data to read from an experiment with e.g., 100,000 pixels is only 391 kB, for which the read and access time with a regular hard drive is <10 ms.

Evaluation Process in msIQuant – from ROI Creation to Evaluation Data Export One of the most important functions of msIQuant for quantitative data analysis is the ability to create and annotate square, polygonal or elliptical ROIs in a tissue section. Elliptical ROIs are often used in quantitation experiments for the annotation of a standard spotted on control tissue. To facilitate annotation of standards, a preset elliptical ROI with defined dimensions can be used. The information and properties given for each ROI are the number of pixels it con-tains, average intensity per pixel, its area, perimeter, volume, and weight. An ROI is created by accessing the image view, in which the ROI can be anno-tated and edited. Read-outs from the ROI can subsequently be used for statis-tical analysis, and creation of calibration standard curves for quantitation.

The quantitation procedures in msIQuant are primarily designed for abso-lute quantitation. However, it can also be used for relative quantitation by de-fining a particular ROI as the reference, then relating intensities of ions de-tected in other ROIs to the intensity in this tissue region. Absolute quantitation is based on standards of known amounts of the compound(s) of interest spot-ted on control tissue. These standards need to be applied in an exact and re-producible way to obtain reliable data. If an endogenous substance is to be quantified, standards should preferably be isotopically labeled substances to avoid the endogenous substances interfering with the resulting standard curve.10 During the sample preparation, an even layer of internal standard should be applied to minimize pixel-to-pixel variation22,24,28 (Figure 3). If it is not possible to apply internal standards, the dataset should be normalized us-ing previously mentioned normalization methods. The advantage of using a deuterated internal standard has been further demonstrated in a comparison of MALDI-MSI and LC-MS/MS quantification results, which deviated by <10%.23,107

41

Figure 12. Quantitation software protocol using msIQuant (Figure from Paper II), illustrated by analysis of a brain tissue section from a drug-administered animal. (A) Ion image of the drug from a sagittal mouse brain tissue section and (B) ion image of an isotope labeled internal standard applied to the tissue section before matrix ap-plication. (C) Normalization of the drug against the internal standard. (D) Drug standards are applied to a control tissue by spotting with different concentrations to form a standard curve. (E) A calibration standard curve is automatically created us-ing the data for the normalized calibration standards. (F) Three brain regions—cere-bral cortex (CTX), hippocampus (HIP), and caudate putamen (CPu)—are selected as ROIs.134 The drug tissue concentration (pmol·mg-1) in each ROI is calculated using msIQuant’s evaluation function.

The quantitation procedure in msIQuant is illustrated in Figure 12, by the quantitation of an administered drug in a sagittal brain tissue section of an experimental animal. Analytical and internal standards are selected and anno-tated from the experimental average spectrum in spectra view. Normalization of the analyte against an internal standard is then selected via image view. From the m/z distribution display, normalization factor distribution is selected to display the normalized distribution of the analyte and the internal standard. ROIs for standards, spotted on control tissue, are then created and annotated. Via image view, a quantitation entry is created where the standard curve is created and displayed simultaneously. The quantitation distribution can then be selected to display the amount of the substance per pixel.

When ROIs have been created and annotated, and the analyte has been se-lected (the source for the analyte data may be an m/z distribution, internal standard-normalized data, or a quantitation distribution), the evaluation func-tion in msIQuant can be obtained via the image view.

The evaluation function generates a list summarizing all the samples in-cluded in the focal project and their ROIs, along with a numerical analysis of the selected ion distribution. The information obtained for each ROI is listed in Table 3. Other information listed in the evaluation module includes the source of the intensity, normalization method, tissue density and thickness. All information generated in an evaluation can be exported by a copy-and-

42

paste functionality into a third-party spreadsheet for further analysis, e.g., sta-tistical analyses.

Table 3. Information provided by the evaluation function of msIQuant (Table from Paper II, supplementary information).

ROI data and values

Tissue name

ROI name

Number of pixels in ROI

ROI Surface area

ROI Perimeter

ROI Volume

ROI Mass

Sum of all intensities in ROI

Intensity sum per area in ROI

Intensity sum per volume in ROI

Intensity sum per mass in ROI

Average value per pixel in ROI

Standard deviation per pixel in ROI

Relative standard deviation per pixel in ROI

Median intensity of pixels in ROI

1st Quartile intensity of pixels in ROI

3d Quartile intensity of pixels in ROI

Minimum intensity of pixels in ROI

Maximum intensity of pixels in ROI

Data Entropy Reduction and Compression of Spectra Data A method for compressing raw MSI datasets before import into msIQuant software was required due to the massive increases in their size following ad-vances in MSI that have improved both spatial and mass resolution. MSI in-struments that are capable of high mass resolution and high spatial resolution may generate several hundred GB of data from an analysis of a single large tissue section (>100,000 pixels). The software packages used to generate MSI raw data in Study III (flexImaging version 4.1 and MassLynx version 4.1)

43

both use un-documented lossless compression methods, since the software structures are proprietary information. Since msIQuant data consist of both spectra and transposed data, a compression algorithm should ideally reduce the data at least sufficiently to correspond to the amount of compressed raw data generated from the instrumental software.

Data in the standardized imzML format35 can be compressed with a method called zlib,135 and OpenMSI software34 can compress data by gzip compres-sion.136 Both these compression methods are based on a lossless technique called DEFLATE.137 However, the Lempel–Ziv–Markov chain algorithm (LZMA)125 provides better compression ratios than DEFLATE, although they are both based on the Lempel–Ziv 1977 (LZ77) algorithm.126 A lossless com-pression algorithm cannot perform any compression on a dataset with truly random numbers because of the Shannon limit,138 but in a mass spectrum there are repetitive sequences, enabling lossless compression. Nevertheless, if a spectrum is overloaded with noise, the compression ratio will be low and the data entropy relatively high.138 There are also lossy compression algorithms, e.g., wavelet transforms that can strongly compress mass spectrometry spec-tra, but they will remove information about noise and the results will be like smoothing and denoising of the spectra.139

One way to increase lossless compression ratios and still conserve infor-mation about the noise is to reduce the data entropy. Thus, a data entropy re-duction method was developed, based on reducing the number of significant figures in the intensity data, and introduced in Paper III. To clarify, an inten-sity array obtained from an MSI experiment may be expressed with less sig-nificant figures since an intensity signal may vary from zero to 107 – 109 counts. For example, if an ion intensity has the value of 9.813854 × 107 and this value is represented by a binary value, the exponent can be expressed with 3 bits, while the mantissa needs to be expressed with 24 bits, a total of 27 bits. If the same ion intensity can be expressed with acceptable accuracy as 9.81 × 107, or even 9.8 × 107, then the mantissa can be expressed with 7 bits and the exponent with 3 bits, so the total size is now 10 bits. This process of reducing the number of bits of a binary value from 27 to 10 bits is the same as reducing the data entropy. The data entropy reduction procedure introduced in Paper III is based on trans-forming the intensity to the natural logarithm and then type casting the loga-rithm from a floating point to an integer data type. The level of reduction has four settings: coarse (10-bits), medium (13-bits), fine (16-bits), and superfine (20-bits). The data entropy reduction and compression algorithm is imple-mented spectrum-by-spectrum in the MSI dataset and pixel-vector-by-pixel-vector in the transposed spectra. This still enables fast reading of a pixel-vec-tor, which corresponds to a specific m/z value. To retrieve the intensity from the compressed pixel-vector, the opposite procedure to the compression is per-

44

formed. By inflating the compressed pixel-vector and executing an exponen-tial operation on the logarithmic value per pixel, the intensity-vector is re-trieved.

Four MSI datasets produced with different ion sources and mass analyzers were investigated (Paper III). The results show that uncompressed msIQuant data can be compressed at ratios ranging from 4.2:1 to 55:1 using the coarse accuracy level. In comparison to raw data generated from commercial soft-ware, the compression ratio at the coarse accuracy level was found to range from 0.9:1 to 5.9:1. Moreover, there was no obvious dataset access time dif-ference between the compressed and uncompressed data formats, e.g., when selecting an m/z value for image visualization, the coarse accuracy compres-sion increased the time required to update the image on the computer screen by 22 ms (24%), from an average of 92 (±32) to 114 (±13) ms (measured using the timer in msIQuant). No clear visible differences were detected between ion distribution images obtained with different levels of data entropy reduc-tion, and image similarity ratios of 99.61% and 99.77% between two uncom-pressed datasets and the respective datasets following compression with coarse accuracy (Figure 13).

We also studied effects of data compression on TIC-normalized intensities of signals showing the distribution of a selected analyte (dopamine) in a cor-onal rat brain tissue section, as detected by MALDI FTICR (Figure 14). The uncompressed dataset gave average intensities (%) of 66.332, 13.290 and 59.616 in the caudate putamen (CPu), anterior commissure, anterior part (aca), and accumbens nucleus, shell region (AcbSh), respectively. The coarse accu-racy compression level gave corresponding average intensities of 66.330, 13.261 and 59.622, respectively, and (hence) deviations from the uncom-pressed dataset of -0.004%, -0.223%, and 0.010%, respectively (less than 0.2% in absolute values).

45

Figure 13. Image difference analysis. (Figure from Paper III, supporting infor-mation). Analysis of differences between images generated from an entropy-reduced and compressed dataset at the four levels and images generated from the uncom-pressed dataset. To perform the difference analysis and exclude bias from image in-terpolation and rainbow color scales, a black and white (B/W) image is analyzed with nearest neighbor interpolation. The differences are derived by subtracting the pixel grayscale values of the uncompressed dataset from those of the compressed da-taset. Positive differences are indicated by the red scale and negative differences by the blue scale. The root mean square (RMS) of the differences and a similarity ratio (SR) are calculated for each compression level. The SR is calculated as 1-RMS/256 expressed as percent. In theory, the pixel difference can occupy 256 discrete levels, so the RMS is divided by 256. (A) DESI dataset: A coronal rat brain tissue section displaying dipalmitoylphosphatidylcholine (DPPC) [M+K]+, m/z 772.5 at a spatial resolution of 150 µm. The black and white color scale shows the intensity at levels from 0 to 255 (white most intense). The image was generated using a DESI ion source coupled to a Q-TOF instrument. (B) FTICR dataset: A coronal rat brain tis-sue section displaying dopamine (DA) derivatized with 2,4-diphenylpyranyl tetra-fluoroborate (DPP-TFB), m/z 368.165, at a spatial resolution of 150 µm. The black and white color scale shows the intensity at levels from 0 to 255. The blue-red devi-ation color scale shows the deviation between the compressed and uncompressed image, e.g., if a pixel from the compressed image has the value 223 and the corre-sponding uncompressed pixel the value 224, the deviation value is -1. The image was generated with a MALDI FTICR mass spectrometer.

46

Figure 14. Quality comparisons of images extracted from (A) an uncompressed da-taset and (B) the dataset compressed at coarse accuracy (10 bit) of dopamine (DA) derivatized with DPP-TFB, m/z 368.165, in a coronal rat-brain-tissue section (modi-fied Figures from Paper III), acquired at a spatial distribution of 150 µm using a MALDI FTICR mass spectrometer. (C) Average deviations in intensity in the im-ages extracted from the two datasets of three brain regions (caudate putamen (CPu), anterior commissure, anterior part (aca) and accumbens nucleus, shell region AcbSh; -0.004%, -0.223%, and -0.010%, respectively).131 The blue-red rainbow color scale represents the intensity from 0 to 100%.

Relative and Absolute Quantitation of Neurotransmitters in Brain Tissue Sections by MALDI Most neurotransmitters cannot be detected by MALDI-MSI due to their poor ionization efficiency. Another concern is overlapping intensities from isobaric fragments and cluster ions from the matrix that interfere and mask detectable signals. To avoid these problems, we developed a procedure to derivatize neu-rotransmitters to improve their ionization efficiency, using pyrylium salts, which react selectively with primary amines at room temperature, ambient pressure and pH > 7 (Paper IV). The reaction products are N-alkyl- or N-aryl-pyridinium derivatives, which are positively charged.115,116 The pyrylium salt used for the derivatization was DPP-TFB (which results in a 215.086 Da mass shift). The DPP-TFB derivatives of endogenous primary amine-containing compounds in a tissue section, including neurotransmitters and their metabo-lites and precursors, undergo self-assisted laser desorption ionization, thus no further matrix is needed. This derivatization strategy enabled the MALDI-MSI detection of derivatized neurotransmitters, metabolites and small endog-enous molecules such as DA, 3-MT, 5-HT, GABA, tyrosine (Tyr), tryptamine, tyramine, phenethylamine and glutamate (Glu). Another pyrylium salt, TMP-TFB (which causes a 105.070 Da mass shift), was used as a confirmatory re-agent.

Some important neurotransmitters, such as ACh and alpha-GPC, do not contain any primary amine and thus do not react with pyrylium cations. A common matrix to analyze small molecules is CHCA, but the fragment peak from CHCA [M-CO2+H]+ (m/z 146.060) overlaps with the signal from ACh (m/z 146.118).

47

Figure 15. Relative quantitation of DA and ACh (modified Figures from Paper IV). The images and bar chart (A-D) show relative quantitation of DA derivatized by DPP-TFB. MS data were acquired using a MALDI-FTICR mass spectrometer at a spatial resolution of 150 µm. (A) Five brain regions were investigated: caudate puta-men (CPu), accumbens nucleus (Acb), cingulate cortex (area 24) (A24), piriform cortex (Pir), medial septal nucleus (MS), and nucleus of the vertical limb of the di-agonal band (VDB).131 (B) High levels of DA were found in the CPu and Acb. (C) By rescaling the signal intensities, it was also possible to detect DA in other parts of the brain. (D) Relative intensities of DA in each region. The inset bars show the re-scaled DA signal intensities for Pir, MS/VDB, and A24. The images and bar chart (E-G) show the substructures of hippocampus and the relative quantitation of ACh. MS images were acquired using a MALDI-TOF/TOF mass spectrometer at a spatial resolution of 15 µm using CHCA-D4 as the matrix. (E) Substructures of the hippo-campus (HIP) were delineated by high-resolution imaging of the distribution of a li-pid with m/z 721.2. (F) The relative abundance and distribution of ACh in the HIP. (G) Relative intensities of ACh signals in four selected HIP sub-regions. High ACh concentrations are apparent in DG-sg and sp regions. Subregions of HIP: granule cell layer of the dentate gyrus (DG-sg), pyramidal layer of CA1 (sp), subiculum (SUB), stratum lacunosum-moleculare (slm), stratum radiatum (sr), stratum oriens (so), cornu ammonis (field 2) (CA2), cornu ammonis (field 3) (CA3), polymorph layer of dentate gyrus (PO).131

However, the problem was solved by synthesizing and applying deuterated CHCA-D4 as the MALDI matrix,108 which shifts the overlapping peak by 4 Da, thus revealing the Ach peak.

msIQuant software was used for both relative and absolute quantitation of GABA, DA, ACh and alpha-GPC. Absolute quantitation of GABA was per-formed on coronal tissue sections of control rat brain, as follows. A standard

48

curve with strong linearity (coefficient of determination, R2 = 0.96) was ob-tained by spotting GABA-D6 standards on a control rat brain section. Average GABA concentrations in the entire tissue section and the medium septum/di-agonal band (MSBD) region, determined using the calibration standard curve, were 5 and 10 nmol/mg, respectively, in line with previous results.140

Absolute quantitation of DA was performed on intact striatum and 6-OHDA lesioned side of a coronal rat brain tissue section. A standard curve with a coefficient of determination of R2 = 0.98 was obtained from spotting DA-D4 standards on control tissue, and concentrations of DA, determined us-ing the curve, in the intact striatum and 6-OHDA lesioned side of the brain were 124 and 84 pmol/mg, respectively. These levels are consistent with pre-vious results obtained by high-performance liquid chromatography (HPLC).141

Relative levels of DA were also determined in five regions of coronal con-trol rat brain (CPu, Acb, Pir, MS & VDB and A24). The highest mean intensity (in Acb) was set to 1 and the intensities in the other regions were expressed proportionally to this level (Figure 15 A-D).

Relative quantitation of ACh was performed with high spatial resolution imaging (15 μm) of the hippocampal area of a sagittal control rat brain tissue section (Figure 15 E-G). The highest concentration of ACh was found in the granule cell layer of the dentate gyrus (DG-sg) and the pyramidal layer of CA1 (sp).

Absolute quantitation of ACh and relative quantitation of alpha-GPC were performed on sagittal mouse brain sections from a control animal and an ani-mal administered with 10 mg/kg of the cholinesterase inhibitor tacrine. The absolute quantitation was based on a calibration standard curve generated by spotting ACh-D9 standards on control tissue. Tacrine administration induced a 7-fold increase in ACh levels in the whole tissue section (Figure 16).

49

Figure 16. MALDI-MS images of ACh and alpha-GPC concentrations in sagittal mouse brain sections from a control animal and after administration of 10 mg/kg of the cholinesterase inhibitor tacrine (Figure from Paper IV). (A) Ion distribution of ACh in the control animal. (B) Ion distribution of ACh in the tacrine-dosed animal (seven times stronger signals, overall, than those from the control animal). (C) Ion image of alpha-GPC from the control animal. (D) Ion image of alpha-GPC from the tacrine-dosed animal (weaker signals than those from the control animal). (E and F). Bar graphs showing the absolute concentration of ACh (E) and relative concentra-tion of alpha-GPC (F) with (black bars) and without (white bars) administration of tacrine in whole brains and selected structures. MS images were acquired using a MALDI-TOF/TOF mass spectrometer at a spatial resolution of 100 µm. Caudate pu-tamen (CPu), hippocampus (HIP), cerebral cortex (CTX), thalamus (TH).134

Relative and Absolute Quantitation of Neurotransmitters and Neuroactive Substances in Brain Tissue Sections by DESI Paper V presents a novel approach for using DESI-MSI to image multiple neurotransmitters, their metabolites and neuroactive drugs directly in brain

50

tissue sections. In this work, msIQuant was used to generate all the images, including the relative and absolute quantitation involved.

When performing DESI-MSI in negative ion mode on a rat brain coronal tissue section, without any further sample preparation, the following neuro-transmitters were detected: adenosine (m/z 302.0667), aspartate (m/z 132.0304), glutamate (m/z 146.0461), glutamine (m/z 145.0621), DA (m/z 152.0719), DA-D3 (m/z 155.0908), and DOPAC (m/z 167.0352). In pos-itive ion mode, the neurotransmitter GABA (m/z 104.0709) was detected.

In Study V, the same strategy for derivatizing substances with primary amines was applied as in Study IV. Derivatization increased the signal/noise ratio for GABA about 100-fold. When investigating coronal rat brain tissue sections from a 6-OHDA lesioned animal and a 6-OHDA lesioned animal ad-ministered with LDOPA-D3, the following DPP-TMP derivatized substances were mapped: GABA (m/z 318.1483), serotonin (m/z 391.1799), DA (m/z 368.1639), DA-D3 (m/z 371.1828), and 3-MT (m/z 382.1795).

To investigate the feasibility of using DESI-MSI to map distributions of administered drugs, tissue sections from amphetamine (m/z 350.1896) admin-istered mice were used. Sagittal mouse brain tissue sections from controls and animals administered amphetamine at concentrations of 1 and 5 mg·kg-1 were analyzed by DESI-MSI. Without derivatization with DPP-TFB, the detected amphetamine ion intensity was just above background noise. However, fol-lowing derivatization, the ion distribution of amphetamine was clearly ob-served and enhanced by a factor of 100. The metabolite 4-hydroxy ampheta-mine (m/z 366.1845) was also detected and mapped in the tissues from the drug-administered animals. Relative quantitation showed that the overall av-erage signal was about six times higher in brain sections from the 5 mg·kg-1

dosed animal than in sections from the 1 mg·kg-1 dosed animal.

51

Figure 17. Relative abundance and distribution of fluvoxamine, determined by DESI-MSI, in a DPP-TFB derivatized sagittal brain tissue section (Figure modified from Pa-per V). (A) Ion distribution of derivatized fluvoxamine (m/z 533.2400) visualized in a tissue section from a fluvoxamine-administered animal (scaled to 40% of maximum intensity). (B) 0.2 μl portions of fluvoxamine solution (serially diluted, at 5:1 dilution ratios, with concentrations ranging from 0.02 mg/ml to 0.032 μg/ml) were spotted on a control brain tissue section (scaled to 5% of maximum intensity). (C) Calibration standard curve for quantitation. (D) Fluvoxamine was quantified in indicated brain structures (pmol/mg) after extracting an average spectrum for each structure. Ion inten-sities were normalized against the internal standard (fluvoxamine-D4). Scale bar, 2 mm; spatial resolution, 100 μm. Cerebellum (Cb), cerebral cortex (CTX), caudate putamen (CPu), ventral striatum (VS), thalamus (TH), hippocampus (HIP).134

Absolute quantitation of the selective serotonin-reuptake-inhibitor fluvoxam-ine was performed on sagittal mouse brain sections from an animal adminis-tered with 40 mg/kg of the drug, using a calibration standard curve obtained by spotting fluvoxamine standards on control tissue. An even layer of internal standard was applied to the tissues using an automatic spraying device (TM-Sprayer) before derivatization with DPP-TFB. The results revealed high con-centration of the compound in several brain regions, e.g., the ventral striatum (Figure 17).

52

Conclusions

Recent developments in MSI have resulted in needs for powerful and func-tional MSI software capable of both qualitative and quantitative data analysis. The functionality of the MSI software msIQuant, developed in the work un-derlying this thesis, has shown great potential to meet these requirements. In Study I, novel MSI software including protocols for quantifying drugs was developed. The software enables key steps in MSI data analysis, such as spec-tra processing, normalization, and quantification. It also offers three normali-zation methods (TIC, RMS, and normalization against internal standards), ef-fects of which were evaluated and compared. The software facilitated quanti-tative data processing, and appeared to be an efficient tool to quantify drugs and endogenous compounds in tissue regions of interest. However, it had some limitations, which prompted improvements detailed in Paper II. Paper II presents the fully developed MSI software msIQuant; an instrument- and manufacturer-independent package that uses the open access MSI data format imzML. The software has been primarily designed for quantitative analyses, but can also be used for qualitative analyses. A user-friendly inter-face was developed that enables easy access and visualization of data, and several features are available for advanced visualization, such as choices of interpolation methods, various color scales for visualizing ion intensities, and possibilities to display data on both linear and logarithmic intensity scales. Functionalities for exporting data, images and data evaluation were also im-plemented. Paper III describes an effective data compression method, based on reducing data entropy by reducing the number of significant figures of the intensity data, followed by compression of the data with the lossless compression method LZMA. Our compression method can be set at four levels: coarse, medium, fine or superfine. We demonstrated that even the largest degree of compression has very little effect on the quality and accuracy of the data. In Study IV, we developed a method to derivatize molecules containing pri-mary amines with pyrylium cations that enables in situ ionization and imaging of neurotransmitters by MALDI-MSI. With this method, we were able to im-

53

age a large number of primary amine-containing neurotransmitters. Concen-trations of the neurotransmitters GABA and DA were quantified directly in tissue sections using deuterated calibration standards. In addition, a deuterated MALDI CHCA-D4 matrix was synthesized, which enabled quantification of the neurotransmitter ACh and its precursor alpha-GPC. Finally, in Paper V, we show that the derivatization method presented in Paper IV improved DESI-MSI, allowing us to quantify neurotransmitters and drugs like amphetamine and the selective serotonin-reuptake-inhibitor fluvoxamine. The improvements enabled detection of low-abundance neurotransmitters and other neuroactive substances previously undetectable by MSI.

54

Future Developments

The current version of msIQuant offers many features for targeted MSI anal-yses and data evaluation, such as imaging of analytes’ distributions, basic sta-tistics, and quantification of single ions. However, msIQuant can only be run on the Windows (x64) platform and many researchers in the MSI research field use Linux and Mac OS X platforms. In addition, for many projects, un-targeted analysis is also required, e.g., in drug discovery projects where there is a need to investigate potential alterations in endogenous substances and me-tabolites by pharmaceutical substances.

To meet these needs the next generation of msIQuant should be able to: Run on Windows, Linux and Mac OS X platforms. An option is to

use the Qt cross platform application framework (qt.io) as the new platform, which would also allow the software to be run on both tab-lets and smartphones.

Run more rapidly, by multi-threading and optimizing processes like spectra processing, conversion from imzML to msIQuant data, and image visualization.

Offer greater ROI functionality by implementing copy and paste ROIs, and enabling abilities to edit and change the nodes/vertices of ROIs.

Display average spectra for each ROI to improve comparisons. Create an assembly and copy experimental regions from many slides

to the assembly, thus enabling comparisons from many experimental runs.

Perform principal component analysis (PCA), and partial least squares (PLS), cluster, receiver operating characteristic (ROC), Bool-ean ROI, and isotope pattern analyses, through incorporation of a sta-tistical package.

Use a cloud storage interface for storing and accessing msIQuant data to enable data sharing in multi-center studies.

55

Populärvetenskaplig sammanfattning

Masspektrometri är en analysteknik som separerar molekyler från varandra och mäter deras molekylvikt. Avbildande masspektrometri (MSI) kallas i dag-ligt tal ’imaging’ masspektrometri. Imaging-tekniken visualiserar hur olika substanser som peptider, proteiner, lipider, läkemedel distribuerar sig i krop-pens olika organ och vävnader, till exempel i hjärnan. Det första experimentet på biologisk vävnad rapporterades 1997 av Richard Caprioli et al. där pepti-ders och proteiners distribution avbildades i ett vävnadssnitt från hjärna. Me-toden har fått allt större betydelse för forskning om sjukdomar och dess orsa-ker, men också inom läkemedelsforskning. Tekniken kan – genom att avbilda molekylerna – spåra vart i kroppen ett läkemedel tar vägen samt även mäta effekter av olika läkemedel. Tekniken är relativt nyutvecklad och tillåter ana-lys av både koncentrationer och det topografiskt ordnade förhållandet av mo-lekyler i en två-dimensionell matris direkt i tunna biologiska vävnadssnitt. Ett mycket stort antal individuella multimodala molekyler kan analyseras samti-digt (>1000) beroende på vilken typ av masspektrometer som används. De vanligaste teknikerna för att jonisera molekyler på biologisk vävnad är MALDI (matrix-assisted laser desorption ionization) eller DESI (desorption electrospray ionization). Avbildning kan utföras med mycket hög molekylär specificitet och med bibehållen anatomisk lokalisation ned till en spatial upp-lösning om c:a 5 µm.

MALDI-MSI tekniken har dock hittills brottats med en rad begränsningar. Det har varit svårt att analysera små molekyler p.g.a. av att den matris som läggs på snittet ger upphov till kluster av joner i det låga masstalsområdet som stör analysen. Matrisen används för att underlätta jonisering av molekyler i väv-nadssnitten. De nya högupplösande MSI instrumenten kan avbilda den spati-ala distributionen och mäta koncentrationer (kvantifiera) av små molekyler direkt i tunna snitt, t ex läkemedel, läkemedelsmetaboliter, neurotransmittorer, lipider, peptider. Två- och tre-dimensionella bilder av små molekylers distri-bution kan för första gången studeras och ge oss nya ledtrådar avseende sjuk-domsorsaker och effekter av behandling.

För att visualisera en specifik molekyls fördelning i ett vävnadssnitt krävs specialiserade programmjukvaror. Mjukvaran msIQuant, som har utvecklats och presenteras i denna avhandling, kan användas för analys av data från MSI experiment (Arbete I, II). Programmet är gratis och kan laddas ned från inter-

56

net. Det unika med mjukvaran är att den är oberoende av tillverkare och in-strumenttyp i och med att den använder ett öppet data format som heter imzML. Genom att omorganisera spektra till dess transponat behöver inte hela datamängden läsas in för vidare analys vilket gör att programmet kan bearbeta stora datamängder utan större fördröjning. Innan visualisering kan ske måste masspektra först processas i olika grad. Därefter normaliseras de med olika metoder för en korrekt visualisering. Både kvalitativa och kvantitativa ana-lyser kan utföras med mjukvaran. Det är möjligt att presentera en molekyls utbredning visuellt och samtidigt beräkna en molekyls koncentration i vävna-den via en kalibreringskurva. Det är även möjligt att definiera och markera specifika vävnadsstrukturer (t ex olika strukturer i hjärnan) och därmed fast-ställa koncentrationen i dessa strukturer. Bilder och mätvärden kan också ex-porteras till andra mjukvaror som kalkyl- och bildredigeringsprogram för ef-terbearbetning och statistiska analyser.

Datamängden som genereras vid MSI experiment är ofta så stor att den inte får plats i datorns internminne vid bearbetning. En viktig funktion i msIQuant är därför att endast en liten del av den totala MSI datan behöver läsas i intern-minnet för vidare analys eftersom den omstruktureras av mjukvaran. En nack-del är dock att datan upptar en stor plats på datorns hårddisk. Dagens högupp-lösande instrument (hög massupplösning och hög spatial upplösning) genere-rar en stor mängd data, ofta mer än 500 GB. För att lösa problemet med stora datamängder har en icke-förstörande komprimeringsfunktion utvecklats i detta avhandlingsarbete som innebär att en mycket hög komprimeringsgrad kan uppnås samtidigt som informationen bibehålls till 99.5% (Arbete III).

Både kvalitativa och kvantitativa studier har utförts med hjälp av msIQuant och exemplifieras i två av de publicerade arbetena i avhandlingen (Arbete IV, V). I det första arbetet utvecklades metoder för att analysera neurotransmitto-rerer och kroppsegna substanser i hjärnvävnadssnitt. Några av neurotransmit-torerna (GABA och dopamin) kunde kvantifieras och vävnadskoncentrationen i hjärnan hos försöksdjur kunde bestämmas. I det andra arbetet analyserades psykoaktiva droger och läkemedelssubstanser i hjärnvävnadssnitt med DESI-MSI. Dels undersöktes den relativa mängden av amfetamin i hjärnan hos för-söksdjur som blivit doserade med olika halter av amfetamin, dels utfördes s.k. absolut kvantifiering av ett SSRI (selektiv serotoninåterupptagshämmare) pre-parat, fluvoxamin. Vid analysen kunde fluvoxaminnivåerna kvantifieras i en-skilda utvalda hjärnregioner. Sammantaget visar resultaten att msIQuant är en kompetent mjukvara som kan användas för både kvalitativa och kvantitativa analyser vid avbildande masspektrometri. Mjukvaran msIQuant har hittills laddats ned av nästan 100 användare från olika världsdelar.

57

Acknowledgements

The work this thesis is based upon was performed at the Biomolecular Mass Spectrometry Imaging (BMSI), Department of Pharmaceutical Biosciences, Uppsala University, Sweden, and was supported by the Swedish Research Council (Medicine and Health grant nos. 2008–5597, 2010–5421, 2011–3170, 2013-3105, Natural and Engineering Science grant no. 2010–5421, 2014-6215, Research Infrastructure 2009-6050), K&A Wallenberg Foundation, Na-tional Institute of Drug Abuse (NIDA, grant no. R21 DA027548-01), the Swe-dish Brain Foundation, the Swedish Foundation for Strategic Research (grant no. RIF14-0078), a Science for Life Laboratory (SciLifeLab) grant, and an Uppsala University Research Infrastructure grant. I would like to express my sincere gratitude to all of you who have contributed to this thesis. In particular, I would like to thank: My supervisor Professor Per Andrén for letting me join his research team, for letting me run the projects with his confidence in me and his great patience during the software development. Great thanks for all support, discussions that resulted in all manuscripts, papers, and finally this thesis. My co-supervisors, Professor Per Svenningsson, Dr. Anna Nilsson and Dr. Mohammadreza Shariatgorji for all the help, discussions, manuscripts writing, support and encouragement. I greatly appreciated all Anna’s and Reza’s help in the lab and with the instruments, and the very fruitful Per and Per annual groups meeting. I would also mention Reza’s magic touch with the chemistry, and that he has been a great chap during all the travels in Europe and overseas, as well as the American jokes. All my colleagues at BMSI, for the all-good times in, and outside, the lab, particularly my roommate and partner in crime, Henrik, whom I have brain-washed with Rammstein and Ozzy. Theodosia always kind and helpful and is great with the statistics. Elva who is the best football player in the lab and the best rapid-fire presenter at young guns. All former co-workers: Michael who contributed a lot to our group, Henrik L who came up with the name msI-Quant, Richard, Sara and Cecilia for all the great discussions. I would also like to mention guest researchers and undergraduate students: Nadine, Bram,

58

Riccardo, Ingo, Lingjie, Jinrui, Florian, and Ásdís for great work and great atmosphere in the lab.

Agneta, Marina, Elisabeth and Martin for all your help with administration of financial issues. Magnus for the last-minute help with poster printing. Lena for always being helpful with various lab issues. Peter and Tobias for helping out with computer and network issues. Mikael O for the help with chemical waste and great chats about hunting and FBK. All colleagues and former colleagues in the B5:2 corridor who contributed to a nice atmosphere and good chats: Maria, Kjell, Ronnie, Ida, Emelie, Tomas, Lisa, Mokhtar, Ahmad, Christine W, Georgy, Tatiana, Igor, Olga, Daniil, Vlad, Anna I, Fred, Mathias, Alfhild, Erika B, Anna J, Erik, Sofia, Jenny, Anna L and Shanti. All colleagues and former colleagues in the department who have provided support and good discussions: Björn H, Eva B, Lennart D, Ingrid N, Margareta H, Ernst O, Myron Z, Raili E, Sigrid E, Stina S, Mikaela H, Johanna E, Mi-chael S, Ola S, Erika R, Anne-Lie S, Lena B, Oskar K, Paula P, Rikard Å, Helén A, Erik B, Anneli W, Jörg H, Inga H, Stina L, Sara P, Shima M, Loudin D, Linnea G, Lova S, and Moustafa M. All colleagues and former colleagues in the Analytical Chemistry Department, who also participated in good discussions: Jonas B, Ingela L, Per S, Konstan-tin A, and Ping S. The “Fika room” for all the great conversations and the weekly fika time! I would also like to thank my friends and former colleagues, with whom I have had a great time and great conversations: Chrille C, Per S, Ulf E, Umeå Ulf E, Gunnar W, Anders T, Owe S, Magdalena A, Lars T, the hunting club Inter-forest, and Odd. I would like to thank my family for all your support, without you, it would have been difficult to complete my doctoral studies. Thank you father and mother, Ulf and Ingegerd for always giving me support and believing in me. Thank you brother with wife, Åke and Maria, my niece Ellen and Viktor. Thank you Maj and Sune for all your support and lovely dinners, Johannes, Jacqueline and Noel. Moa and Lilly – Moa, thank you for all your support and for being my life companion. Especially thanks for all the hard work you have done to help me during the last months when my focus has been completely on my work. Lilly, thank you for being my bonus daughter and for giving me perspective in life.

59

References

1. Caprioli, R.M., T.B. Farmer, and J. Gile, Molecular imaging of biological sam-ples: localization of peptides and proteins using MALDI-TOF MS. Anal Chem 1997, 69, 4751-4760.

2. Woods, A.S. and S.N. Jackson, Brain tissue lipidomics: direct probing using ma-trix-assisted laser desorption/ionization mass spectrometry. AAPS J 2006, 8, E391-395.

3. Touboul, D., et al., Changes of phospholipid composition within the dystrophic muscle by matrix-assisted laser desorption/ionization mass spectrometry and mass spectrometry imaging. European Journal of Mass Spectrometry 2004, 10, 657-664.

4. Gode, D. and D.A. Volmer, Lipid imaging by mass spectrometry - a review. Analyst 2013, 138, 1289-1315.

5. Berry, K.A., et al., MALDI imaging of lipid biochemistry in tissues by mass spectrometry. Chem Rev 2011, 111, 6491-6512.

6. Stoeckli, M., et al., Imaging mass spectrometry: a new technology for the anal-ysis of protein expression in mammalian tissues. Nat Med 2001, 7, 493-496.

7. Skold, K., et al., Decreased striatal levels of PEP-19 following MPTP lesion in the mouse. J Proteome Res 2006, 5, 262-269.

8. Shariatgorji, M., et al., Simultaneous imaging of multiple neurotransmitters and neuroactive substances in the brain by desorption electrospray ionization mass spectrometry. Neuroimage 2016, 136, 129-138.

9. Bergman, H.M., et al., Quantitative mass spectrometry imaging of small-mole-cule neurotransmitters in rat brain tissue sections using nanospray desorption electrospray ionization. Analyst 2016, 141, 3686-3695.

10. Shariatgorji, M., et al., Direct targeted quantitative molecular imaging of neuro-transmitters in brain tissue sections. Neuron 2014, 84, 697-707.

11. Nilsson, A., et al., Mass spectrometry imaging in drug development. Anal Chem 2015, 87, 1437-1455.

12. Nilsson, A., et al., Fine mapping the spatial distribution and concentration of unlabeled drugs within tissue micro-compartments using imaging mass spec-trometry. PLoS One 2010, 5, e11411.

13. Reyzer, M.L., et al., Direct analysis of drug candidates in tissue by matrix-as-sisted laser desorption/ionization mass spectrometry. J Mass Spectrom 2003, 38, 1081-1092.

14. Norris, J.L. and R.M. Caprioli, Analysis of tissue specimens by matrix-assisted laser desorption/ionization imaging mass spectrometry in biological and clinical research. Chem Rev 2013, 113, 2309-2342.

15. Seeley, E.H. and R.M. Caprioli, MALDI imaging mass spectrometry of human tissue: method challenges and clinical perspectives. Trends Biotechnol 2011, 29, 136-143.

16. Lemaire, R., et al., Direct analysis and MALDI imaging of formalin-fixed, par-affin-embedded tissue sections. J Proteome Res 2007, 6, 1295-1305.

60

17. Lietz, C.B., E. Gemperline, and L. Li, Qualitative and quantitative mass spec-trometry imaging of drugs and metabolites. Adv Drug Deliv Rev 2013, 65, 1074-1085.

18. Guenther, S., et al., AP-MALDI imaging of neuropeptides in mouse pituitary gland with 5 mu m spatial resolution and high mass accuracy. Int J Mass Spec-trom 2011, 305, 228-237.

19. Stoeckli, M., et al., Imaging of a beta-peptide distribution in whole-body mice sections by MALDI mass spectrometry. J Am Soc Mass Spectrom 2007, 18, 1921-1924.

20. Wiseman, J.M., et al., Tissue imaging at atmospheric pressure using desorption electrospray ionization (DESI) mass spectrometry. Angew Chem Int Ed Engl 2006, 45, 7188-7192.

21. Rzagalinski, I. and D.A. Volmer, Quantification of low molecular weight com-pounds by MALDI imaging mass spectrometry - A tutorial review. Biochim Bi-ophys Acta 2017, 1865, 726-739.

22. Kallback, P., et al., msIQuant--Quantitation Software for Mass Spectrometry Im-aging Enabling Fast Access, Visualization, and Analysis of Large Data Sets. Anal Chem 2016, 88, 4346-4353.

23. Chumbley, C.W., et al., Absolute Quantitative MALDI Imaging Mass Spectrom-etry: A Case of Rifampicin in Liver Tissues. Anal Chem 2016, 88, 2392-2398.

24. Porta, T., et al., Quantification in MALDI-MS imaging: what can we learn from MALDI-selected reaction monitoring and what can we expect for imaging? Anal Bioanal Chem 2015, 407, 2177-2187.

25. Jadoul, L., et al., A spiked tissue-based approach for quantification of phospha-tidylcholines in brain section by MALDI mass spectrometry imaging. Anal Bio-anal Chem 2015, 407, 2095-2106.

26. Pirman, D.A., et al., Quantitative MALDI tandem mass spectrometric imaging of cocaine from brain tissue with a deuterated internal standard. Anal Chem 2013, 85, 1081-1089.

27. Groseclose, M.R. and S. Castellino, A mimetic tissue model for the quantifica-tion of drug distributions by MALDI imaging mass spectrometry. Anal Chem 2013, 85, 10099-10106.

28. Kallback, P., et al., Novel mass spectrometry imaging software assisting labeled normalization and quantitation of drugs and neuropeptides directly in tissue sec-tions. J Proteomics 2012, 75, 4941-4951.

29. Hamm, G., et al., Quantitative mass spectrometry imaging of propranolol and olanzapine using tissue extinction calculation as normalization factor. J Prote-omics 2012, 75, 4952-4961.

30. Pirman, D.A. and R.A. Yost, Quantitative tandem mass spectrometric imaging of endogenous acetyl-L-carnitine from piglet brain tissue using an internal stand-ard. Anal Chem 2011, 83, 8575-8581.

31. Vismeh, R., et al., Localization and quantification of drugs in animal tissues by use of desorption electrospray ionization mass spectrometry imaging. Anal Chem 2012, 84, 5439-5445.

32. Rompp, A., et al., Mass spectrometry imaging with high resolution in mass and space (HR(2) MSI) for reliable investigation of drug compound distributions on the cellular level. Anal Bioanal Chem 2011, 401, 65-73.

33. Herrmann, A.M., et al., Nano-scale secondary ion mass spectrometry - A new analytical tool in biogeochemistry and soil ecology: A review article. Soil Biol Biochem 2007, 39, 1835-1850.

34. Rubel, O., et al., OpenMSI: a high-performance web-based platform for mass spectrometry imaging. Anal Chem 2013, 85, 10354-10361.

61

35. Rompp, A., et al., imzML: Imaging Mass Spectrometry Markup Language: A common data format for mass spectrometry imaging. Methods Mol Biol 2011, 696, 205-224.

36. Tanaka, K., et al., Protein and polymer analyses up to m/z 100 000 by laser ion-ization time-of-flight mass spectrometry. Rapid Commun Mass Spectrom 1988, 2, 151-153.

37. Karas, M., et al., Matrix-Assisted Ultraviolet-Laser Desorption of Nonvolatile Compounds. Int J Mass Spectrom 1987, 78, 53-68.

38. Karas, M., D. Bachmann, and F. Hillenkamp, Influence of the Wavelength in High-Irradiance Ultraviolet-Laser Desorption Mass-Spectrometry of Organic-Molecules. Anal Chem 1985, 57, 2935-2939.

39. Armichev, A.V., V.S. Aleinikov, and T.B. Fogelson, High-Power Tea Uv Nitro-gen Laser without Gas-Flow. Kvantovaya Elektron+ 1980, 7, 1037-1041.

40. Haase, A. and J. Hoehndorf, Mass spectrometer with MALDI laser system, U.S. Patent No. 8,431,890 2013.

41. Karas, M. and U. Bahr, Laser Desorption Ionization Mass-Spectrometry of Large Biomolecules. TrAC Trends in Analytical Chemistry 1990, 9, 321-325.

42. Beavis, R.C., T. Chaudhary, and B.T. Chait, Alpha-Cyano-4-Hydroxycinnamic Acid as a Matrix for Matrix-Assisted Laser Desorption Mass-Spectrometry. Org Mass Spectrom 1992, 27, 156-158.

43. Juhasz, P. and C.E. Costello, Generation of large radical ions from oligometal-locenes by matrix-assisted laser desorption ionization. Rapid Commun Mass Spectrom 1993, 7, 343-351.

44. Pieles, U., et al., Matrix-assisted laser desorption ionization time-of-flight mass spectrometry: a powerful tool for the mass and sequence analysis of natural and modified oligonucleotides. Nucleic Acids Res 1993, 21, 3191-3196.

45. Demesmaeker, A., et al., Amides as a New-Type of Backbone Modification in Oligonucleotides. Angew Chem Int Edit 1994, 33, 226-229.

46. Vermillion-Salsbury, R.L. and D.M. Hercules, 9-aminoacridine as a matrix for negative mode matrix-assisted laser desorption/ionization. Rapid Commun Mass Spectrom 2002, 16, 1575-1581.

47. Aerni, H.R., D.S. Cornett, and R.M. Caprioli, Automated acoustic matrix depo-sition for MALDI sample preparation. Anal Chem 2006, 78, 827-834.

48. Chen, Y., et al., Imaging MALDI mass spectrometry using an oscillating capil-lary nebulizer matrix coating system and its application to analysis of lipids in brain from a mouse model of Tay-Sachs/Sandhoff disease. Anal Chem 2008, 80, 2780-2788.

49. Cornett, D.S., et al., MALDI imaging mass spectrometry: molecular snapshots of biochemical systems. Nat Methods 2007, 4, 828-833.

50. Groseclose, M.R., et al., Identification of proteins directly from tissue: in situ tryptic digestions coupled with imaging mass spectrometry. J Mass Spectrom 2007, 42, 254-262.

51. Schwartz, S.A., M.L. Reyzer, and R.M. Caprioli, Direct tissue analysis using matrix-assisted laser desorption/ionization mass spectrometry: practical aspects of sample preparation. J Mass Spectrom 2003, 38, 699-708.

52. Walch, A., et al., MALDI imaging mass spectrometry for direct tissue analysis: a new frontier for molecular histology. Histochem Cell Biol 2008, 130, 421-434.

53. Puolitaival, S.M., et al., Solvent-free matrix dry-coating for MALDI imaging of phospholipids. J Am Soc Mass Spectrom 2008, 19, 882-886.

62

54. Goodwin, R.J., et al., Qualitative and quantitative MALDI imaging of the posi-tron emission tomography ligands raclopride (a D2 dopamine antagonist) and SCH 23390 (a D1 dopamine antagonist) in rat brain tissue sections using a sol-vent-free dry matrix application method. Anal Chem 2011, 83, 9694-9701.

55. Hankin, J.A., R.M. Barkley, and R.C. Murphy, Sublimation as a method of ma-trix application for mass spectrometric imaging. J Am Soc Mass Spectrom 2007, 18, 1646-1652.

56. Takats, Z., et al., Mass spectrometry sampling under ambient conditions with desorption electrospray ionization. Science 2004, 306, 471-473.

57. Gerbig, S., et al., Spatially resolved investigation of systemic and contact pesti-cides in plant material by desorption electrospray ionization mass spectrometry imaging (DESI-MSI). Anal Bioanal Chem 2015, 407, 7379-7389.

58. Herzog, R.F.K. and F.P. Viehbock, Ion Source for Mass Spectrography. Phys Rev 1949, 76, 855-856.

59. Honig, R.E., Mass Spectrometer Study of the Sputtering of Surfaces. Anal Chem 1957, 29, 1379-1379.

60. Storms, H.A., K.F. Brown, and J.D. Stein, Evaluation of a Cesium Positive-Ion Source for Secondary Ion Mass-Spectrometry. Anal Chem 1977, 49, 2023-2030.

61. Nygren, H., et al., Localization of cholesterol, phosphocholine and galacto-sylceramide in rat cerebellar cortex with imaging TOF-SIMS equipped with a bismuth cluster ion source. Biochim Biophys Acta 2005, 1737, 102-110.

62. Weibel, D., et al., A C-60 primary ion beam system for time of flight secondary ion mass spectrometry: Its development and secondary ion yield characteristics. Anal Chem 2003, 75, 1754-1764.

63. Fletcher, J.S., N.P. Lockyer, and J.C. Vickerman, C-60, buckminsterfullerene: its impact on biological ToF-SIMS analysis. Surf Interface Anal 2006, 38, 1393-1400.

64. Wittmaack, K., Primary-ion charge compensation in SIMS analysis of insula-tors. Journal of Applied Physics 1978, 52, 493-497.

65. Passarelli, M.K. and N. Winograd, Lipid imaging with time-of-flight secondary ion mass spectrometry (ToF-SIMS). Biochim Biophys Acta 2011, 1811, 976-990.

66. Colton, R.J., et al., High-Performance Secondary Ion Mass-Spectrometer. Rev Sci Instrum 1980, 51, 1685-1689.

67. Vandervorst, W. and F.R. Shepherd, Secondary Ion Mass-Spectrometry Profil-ing of Shallow, Implanted Layers Using Quadrupole and Magnetic-Sector In-struments. J Vac Sci Technol A 1987, 5, 313-320.

68. Hofer, W.O. and F. Thum, Simple Axially-Symmetric Quadrupole Sims Spec-trometer. Nucl Instrum Methods 1978, 149, 535-541.

69. Benninghoven, A., Chemical-Analysis of Inorganic and Organic-Surfaces and Thin-Films by Static Time-of-Flight Secondary-Ion Mass-Spectrometry (Tof-Sims). Angew Chem Int Edit 1994, 33, 1023-1043.

70. Oechsner, H., Subcellular imaging of cells and tissues with dynamic secondary ion mass spectrometry. Plasma Phys 1973, 16, 835-844.

71. Watrous, J.D. and P.C. Dorrestein, Imaging mass spectrometry in microbiology. Nature Rev Microbiol 2011, 9, 683.

72. Roach, P.J., J. Laskin, and A. Laskin, Molecular characterization of organic aer-osols using nanospray-desorption/electrospray ionization-mass spectrometry. Anal Chem 2010, 82, 7979-7986.

73. Laskin, J., et al., Tissue imaging using nanospray desorption electrospray ioni-zation mass spectrometry. Anal Chem 2012, 84, 141-148.

63

74. Wiley, W.C. and I.H. Mclaren, Time-of-Flight Mass Spectrometer with Im-proved Resolution. Rev Sci Instrum 1955, 26, 1150-1157.

75. Prentice, B.M. and R.M. Caprioli, The Need for Speed in Matrix-Assisted Laser Desorption/Ionization Imaging Mass Spectrometry. Postdoc J 2016, 4, 3-13.

76. Chernushevich, I.V., A.V. Loboda, and B.A. Thomson, An introduction to quad-rupole-time-of-flight mass spectrometry. J Mass Spectrom 2001, 36, 849-865.

77. Comisarow, M.B. and A.G. Marshall, Fourier-Transform Ion-Cyclotron Reso-nance Spectroscopy. Chem Phys Lett 1974, 25, 282-283.

78. Marshall, A.G., Fourier-Transform Ion-Cyclotron Resonance Mass-Spectrome-try. Acc Chem Res 1985, 18, 316-322.

79. Marshall, A.G., C.L. Hendrickson, and G.S. Jackson, Fourier transform ion cy-clotron resonance mass spectrometry: a primer. Mass Spectrom Rev 1998, 17, 1-35.

80. Hardman, M. and A.A. Makarov, Interfacing the orbitrap mass analyzer to an electrospray ion source. Anal Chem 2003, 75, 1699-1705.

81. Hu, Q., et al., The Orbitrap: a new mass spectrometer. J Mass Spectrom 2005, 40, 430-443.

82. Makarov, A., Mass spectrometer, U.S. Patent No. 5,886,346 1999. 83. Makarov, A., Electrostatic axially harmonic orbital trapping: a high-perfor-

mance technique of mass analysis. Anal Chem 2000, 72, 1156-1162. 84. Norris, J.L., et al., Processing MALDI Mass Spectra to Improve Mass Spectral

Direct Tissue Analysis. Int J Mass Spectrom 2007, 260, 212-221. 85. Garden, R.W. and J.V. Sweedler, Heterogeneity within MALDI samples as re-

vealed by mass spectrometric imaging. Anal Chem 2000, 72, 30-36. 86. Luxembourg, S.L., et al., Effect of local matrix crystal variations in matrix-as-

sisted ionization techniques for mass spectrometry. Anal Chem 2003, 75, 2333-2341.

87. Deininger, S.O., et al., Normalization in MALDI-TOF imaging datasets of pro-teins: practical considerations. Anal Bioanal Chem 2011, 401, 167-181.

88. Barry, J.A., G. Robichaud, and D.C. Muddiman, Mass recalibration of FT-ICR mass spectrometry imaging data using the average frequency shift of ambient ions. J Am Soc Mass Spectrom 2013, 24, 1137-1145.

89. Gobom, J., et al., A calibration method that simplifies and improves accurate determination of peptide molecular masses by MALDI-TOF MS. Anal Chem 2002, 74, 3915-3923.

90. Kozhinov, A.N., K.O. Zhurov, and Y.O. Tsybin, Iterative method for mass spec-tra recalibration via empirical estimation of the mass calibration function for Fourier transform mass spectrometry-based petroleomics. Anal Chem 2013, 85, 6437-6445.

91. Kulkarni, P., et al., Correcting mass shifts: A lock mass-free recalibration pro-cedure for mass spectrometry imaging data. Anal Bioanal Chem 2015, 407, 7603-7613.

92. Stravs, M.A., et al., Automatic recalibration and processing of tandem mass spectra using formula annotation. J Mass Spectrom 2013, 48, 89-99.

93. Wolski, W.E., et al., Calibration of mass spectrometric peptide mass fingerprint data without specific external or internal calibrants. BMC Bioinform 2005, 6, 203.

94. Rauch, M. and M. Stoeckli, Biomap 2000, https://ms-imaging.org/wp/biomap/. 95. Biomedical Imaging Resource, Mayo Clinic, Rochester, MN, USA, ANALYZE

7.5 File Format 1998, https://rportal.mayo.edu/bir/ ANALYZE75.pdf. 96. Giesen, C., et al., Highly multiplexed imaging of tumor tissues with subcellular

resolution by mass cytometry. Nat Methods 2014, 11, 417-422.

64

97. Gustafsson, J.O., et al., Internal calibrants allow high accuracy peptide matching between MALDI imaging MS and LC-MS/MS. J Proteomics 2012, 75, 5093-5105.

98. Robbe, M.F., et al., Software tools of the Computis European project to process mass spectrometry images. European Journal of Mass Spectrometry 2014, 20, 351-360.

99. Klinkert, I., et al., Methods for full resolution data exploration and visualization for large 2D and 3D mass spectrometry imaging datasets. Int J Mass Spectrom 2014, 362, 40-47.

100. Paschke, C., et al., Mirion--a software package for automatic processing of mass spectrometric images. J Am Soc Mass Spectrom 2013, 24, 1296-1306.

101. Robichaud, G., et al., MSiReader: an open-source interface to view and analyze high resolving power MS imaging files on Matlab platform. J Am Soc Mass Spectrom 2013, 24, 718-721.

102. Parry, R.M., et al., omniSpect: an open MATLAB-based tool for visualization and analysis of matrix-assisted laser desorption/ionization and desorption elec-trospray ionization mass spectrometry images. J Am Soc Mass Spectrom 2013, 24, 646-649.

103. Bemis, K.D., et al., Cardinal: an R package for statistical analysis of mass spec-trometry-based imaging experiments. Bioinformatics 2015, 31, 2418-2420.

104. Race, A.M., et al., SpectralAnalysis: Software for the Masses. Anal Chem 2016, 88, 9451-9458.

105. Corr, J.J., et al., Design considerations for high speed quantitative mass spec-trometry with MALDI ionization. J Am Soc Mass Spectrom 2006, 17, 1129-1141.

106. Koeniger, S.L., et al., A quantitation method for mass spectrometry imaging. Rapid Commun Mass Spectrom 2011, 25, 503-510.

107. Prentice, B.M., C.W. Chumbley, and R.M. Caprioli, Absolute Quantification of Rifampicin by MALDI Imaging Mass Spectrometry Using Multiple TOF/TOF Events in a Single Laser Shot. J Am Soc Mass Spectrom 2017, 28, 136-144.

108. Shariatgorji, M., et al., Deuterated matrix-assisted laser desorption ionization matrix uncovers masked mass spectrometry imaging signals of small molecules. Anal Chem 2012, 84, 7152-7157.

109. de Jong, L.A., et al., Receptor-ligand binding assays: technologies and applica-tions. J Chromatogr B Analyt Technol Biomed Life Sci 2005, 829, 1-25.

110. Falck, B., et al., Fluorescence of Catechol Amines and Related Compounds Con-densed with Formaldehyde. J Histochem Cytochem 1962, 10, 348-&.

111. Jones, B.E. and A. Beaudet, Distribution of acetylcholine and catecholamine neurons in the cat brainstem: a choline acetyltransferase and tyrosine hydrox-ylase immunohistochemical study. J Comp Neurol 1987, 261, 15-32.

112. Keenan, C.L. and H. Koopowitz, Limitations in identifying neurotransmitters within neurons by fluorescent histochemistry techniques. Science 1981, 214, 1151-1152.

113. Pimlott, S.L. and A. Sutherland, Molecular tracers for the PET and SPECT im-aging of disease. Chem Soc Rev 2011, 40, 149-162.

114. Badgaiyan, R.D., Neurotransmitter Imaging: Basic Concepts and Future Per-spectives. Curr Med Imaging Rev 2011, 7, 98-103.

115. Katritzky, A.R., et al., Mechanism of the Pyrylium-Pyridinium Ring Intercon-version - Mild Preparative Conditions for Conversion of Amines into Pyridinium Ions. Angew Chem Int Edit 1980, 19, 306-306.

116. Katritzky, A.R. and S.S. Thind, The Synthesis and Reactions of Sterically Con-strained Pyrylium and Pyridinium Salts. J Chem Soc Perk T 1 1980, 1895-1900.

65

117. Fernagut, P.O., et al., Dopamine transporter binding is unaffected by L-DOPA administration in normal and MPTP-treated monkeys. PLoS One 2010, 5, e14053.

118. Porras, G., et al., PSD-95 expression controls L-DOPA dyskinesia through do-pamine D1 receptor trafficking. J Clin Invest 2012, 122, 3977-3989.

119. Santini, E., et al., Distinct changes in cAMP and extracellular signal-regulated protein kinase signalling in L-DOPA-induced dyskinesia. PLoS One 2010, 5, e12322.

120. Bezard, E., et al., A chronic MPTP model reproducing the slow evolution of Parkinson's disease: evolution of motor symptoms in the monkey. Brain Res 1997, 766, 107-112.

121. Bezard, E., et al., Relationship between the appearance of symptoms and the level of nigrostriatal degeneration in a progressive 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine-lesioned macaque model of Parkinson's disease. J Neurosci 2001, 21, 6853-6861.

122. Bezard, E., et al., Upregulation of striatal preproenkephalin gene expression oc-curs before the appearance of parkinsonian signs in 1-methyl-4-phenyl- 1,2,3,6-tetrahydropyridine monkeys. Neurobiol Dis 2001, 8, 343-350.

123. Savitzky, A. and M.J.E. Golay, Smoothing and Differentiation of Data by Sim-plified Least Squares Procedures. Anal Chem 1964, 36, 1627-&.

124. McDonnell, L.A., et al., Imaging mass spectrometry data reduction: automated feature identification and extraction. J Am Soc Mass Spectrom 2010, 21, 1969-1978.

125. Pavlov, I., LZMA SDK (Software Development Kit) 1998, http://www.7-zip.org/.

126. Ziv, J. and A. Lempel, A Universal Algorithm for Sequential Data Compression. IEEE Trans Inf Theory 1977, 23, 337-343.

127. Choi, M., A new intensity-hue-saturation fusion approach to image fusion with a tradeoff parameter. Ieee T Geosci Remote 2006, 44, 1672-1682.

128. Daneshvar, S. and H. Ghassemian, MRI and PET image fusion by combining IHS and retina-inspired models. Inform Fusion 2010, 11, 114-123.

129. He, C.T., et al., Multimodal medical image fusion based on IHS and PCA. Pro-cedia Engineer 2010, 7, 280-285.

130. Van de Plas, R., et al., Image fusion of mass spectrometry and microscopy: a multimodality paradigm for molecular tissue mapping. Nat Methods 2015, 12, 366-372.

131. Paxinos, G. and C. Watson, The rat brain in stereotaxic coordinates, 7th ed., Else-vier, 2014.

132. McCombie, G., et al., Spatial and spectral correlations in MALDI mass spec-trometry images by clustering and multivariate analysis. Anal Chem 2005, 77, 6118-6124.

133. McDonnell, L.A. and R.M.A. Heeren, Imaging mass spectrometry. Mass Spec-trom Rev 2007, 26, 606-643.

134. Paxinos, G. and K.B.J. Franklin, The mouse brain in stereotaxic coordinates, 4th ed., Elsevier, 2013.

135. Gailly, J.l. and M. Adler, zlib 1995, https://zlib.net/. 136. Gailly, J.l. and M. Adler, gzip 1992, https://www.gnu.org/software/ gzip/. 137. Katz, P.W., String searcher, and compressor using same, U.S. Patent No.

5,051,745 1991. 138. Shannon, C.E., A Mathematical Theory of Communication. At&T Tech J 1948,

27, 379-423.

66

139. Barclay, V.J., R.F. Bonner, and I.P. Hamilton, Application of wavelet transforms to experimental spectra: Smoothing, denoising, and data set compression. Anal Chem 1997, 69, 78-90.

140. Schaaf, J.M., et al., A specific and sensitive determination of gamma aminobu-tyric acid in CSF and brain tissue by gas chromatography-mass spectrometry. J Neurosci Methods 1985, 13, 257-265.

141. Zhao, X.E. and Y.R. Suo, Simultaneous determination of monoamine and amino acid neurotransmitters in rat endbrain tissues by pre-column derivatization with high-performance liquid chromatographic fluorescence detection and mass spec-trometric identification. Talanta 2008, 76, 690-697.

Acta Universitatis UpsaliensisDigital Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Pharmacy 234

Editor: The Dean of the Faculty of Pharmacy

A doctoral dissertation from the Faculty of Pharmacy, UppsalaUniversity, is usually a summary of a number of papers. A fewcopies of the complete dissertation are kept at major Swedishresearch libraries, while the summary alone is distributedinternationally through the series Digital ComprehensiveSummaries of Uppsala Dissertations from the Faculty ofPharmacy. (Prior to January, 2005, the series was publishedunder the title “Comprehensive Summaries of UppsalaDissertations from the Faculty of Pharmacy”.)

Distribution: publications.uu.seurn:nbn:se:uu:diva-328016

ACTAUNIVERSITATIS

UPSALIENSISUPPSALA

2017